From 08223c06a59c3b841d72286ef120494e9cf10c65 Mon Sep 17 00:00:00 2001
From: Aska <Aska@cyrene.local>
Date: Sun, 6 Sep 2026 06:26:59 +0800
Subject: [PATCH] =?UTF-8?q?feat(tools/basic):=20=E5=86=85=E7=BD=AE?=
 =?UTF-8?q?=E5=9F=BA=E7=A1=80=E5=B7=A5=E5=85=B7=EF=BC=88calculator(?=
 =?UTF-8?q?=E6=B3=A8=E5=85=A5=E5=AE=89=E5=85=A8)/datetime/random/text/json?=
 =?UTF-8?q?/markdown=EF=BC=89+=20=E6=89=B9=E9=87=8F=E6=B3=A8=E5=86=8C?=
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

---
 plugins/tools/basic/calculator.go      | 317 +++++++++++++++++++++++++
 plugins/tools/basic/calculator_test.go | 113 +++++++++
 plugins/tools/basic/context_test.go    |  25 ++
 plugins/tools/basic/datetime.go        |  77 ++++++
 plugins/tools/basic/datetime_test.go   |  99 ++++++++
 plugins/tools/basic/doc.go             |   9 +
 plugins/tools/basic/helpers_test.go    | 104 ++++++++
 plugins/tools/basic/json.go            |  95 ++++++++
 plugins/tools/basic/json_test.go       | 119 ++++++++++
 plugins/tools/basic/log.go             |  17 ++
 plugins/tools/basic/markdown.go        | 193 +++++++++++++++
 plugins/tools/basic/markdown_test.go   | 121 ++++++++++
 plugins/tools/basic/random.go          | 125 ++++++++++
 plugins/tools/basic/random_test.go     |  86 +++++++
 plugins/tools/basic/register.go        |  37 +++
 plugins/tools/basic/register_test.go   |  77 ++++++
 plugins/tools/basic/text.go            | 161 +++++++++++++
 plugins/tools/basic/text_test.go       | 107 +++++++++
 plugins/tools/basic/util.go            | 192 +++++++++++++++
 19 files changed, 2074 insertions(+)
 create mode 100644 plugins/tools/basic/calculator.go
 create mode 100644 plugins/tools/basic/calculator_test.go
 create mode 100644 plugins/tools/basic/context_test.go
 create mode 100644 plugins/tools/basic/datetime.go
 create mode 100644 plugins/tools/basic/datetime_test.go
 create mode 100644 plugins/tools/basic/doc.go
 create mode 100644 plugins/tools/basic/helpers_test.go
 create mode 100644 plugins/tools/basic/json.go
 create mode 100644 plugins/tools/basic/json_test.go
 create mode 100644 plugins/tools/basic/log.go
 create mode 100644 plugins/tools/basic/markdown.go
 create mode 100644 plugins/tools/basic/markdown_test.go
 create mode 100644 plugins/tools/basic/random.go
 create mode 100644 plugins/tools/basic/random_test.go
 create mode 100644 plugins/tools/basic/register.go
 create mode 100644 plugins/tools/basic/register_test.go
 create mode 100644 plugins/tools/basic/text.go
 create mode 100644 plugins/tools/basic/text_test.go
 create mode 100644 plugins/tools/basic/util.go

diff --git a/plugins/tools/basic/calculator.go b/plugins/tools/basic/calculator.go
new file mode 100644
index 0000000..33f7f95
--- /dev/null
+++ b/plugins/tools/basic/calculator.go
@@ -0,0 +1,317 @@
+package basic
+
+import (
+	"context"
+	"fmt"
+	"math"
+	"strconv"
+	"strings"
+)
+
+// maxExprLen bounds an expression length so a request cannot force a parse/scan
+// beyond a trivial cost and cannot smuggle a larger payload through the tool.
+const maxExprLen = 256
+
+// allowedChars lists the bytes permitted in an expression besides ASCII letters
+// (letters are only admitted to spell the sqrt function; the lexer rejects any
+// other identifier).
+const allowedChars = "0123456789.+-*/^() \t\n"
+
+// calculator evaluates a safe arithmetic expression. It recognizes signed numeric
+// literals, + - * /, parentheses, the right-associative ^ power operator and the
+// sqrt(x) function. Every other character or identifier is rejected, so the
+// expression can never run arbitrary code. Division by zero and non-real results
+// are errors.
+type calculator struct{}
+
+// Name returns the tool identifier.
+func (calculator) Name() string { return "calculator" }
+
+// Description explains what the tool does for the model.
+func (calculator) Description() string {
+	return "Safely evaluate an arithmetic expression using + - * /, parentheses, ^ (power) and sqrt(x)."
+}
+
+// Parameters returns the tool input JSON Schema.
+func (calculator) Parameters() (map[string]any, error) {
+	return map[string]any{
+		"type": "object",
+		"properties": map[string]any{
+			"expression": map[string]any{"type": "string", "description": "the arithmetic expression to evaluate"},
+		},
+		"required": []any{"expression"},
+	}, nil
+}
+
+// Execute evaluates expression and returns {"expression":..., "result":...}.
+func (calculator) Execute(ctx context.Context, args any) (any, error) {
+	if err := ctx.Err(); err != nil {
+		return nil, err
+	}
+	m, err := asMap(args)
+	if err != nil {
+		return nil, err
+	}
+	expr, err := reqString(m, "expression")
+	if err != nil {
+		return nil, err
+	}
+	val, err := eval(expr)
+	if err != nil {
+		log().Error("calculator failed", "expression", expr, "reason", err.Error())
+		return nil, err
+	}
+	return map[string]any{"expression": expr, "result": val}, nil
+}
+
+// eval validates, tokenizes and parses expr, returning its numeric value.
+func eval(expr string) (float64, error) {
+	if expr == "" {
+		return 0, fmt.Errorf("%w: expression is empty", ErrEvaluation)
+	}
+	if len(expr) > maxExprLen {
+		return 0, fmt.Errorf("%w: expression exceeds %d characters", ErrEvaluation, maxExprLen)
+	}
+	for _, c := range expr {
+		if strings.ContainsRune(allowedChars, c) {
+			continue
+		}
+		if c >= 'a' && c <= 'z' || c >= 'A' && c <= 'Z' {
+			continue // letter: the lexer admits only "sqrt"
+		}
+		return 0, fmt.Errorf("%w: disallowed character %q", ErrEvaluation, c)
+	}
+	toks, err := lex(expr)
+	if err != nil {
+		return 0, err
+	}
+	p := &parser{toks: toks}
+	return p.parse()
+}
+
+func isAsciiLetter(c byte) bool {
+	return c >= 'a' && c <= 'z' || c >= 'A' && c <= 'Z'
+}
+
+// tokKind describes a token category produced by the lexer.
+type tokKind int
+
+const (
+	tokEOF tokKind = iota
+	tokNum
+	tokOp
+	tokLParen
+	tokRParen
+	tokIdent
+)
+
+type token struct {
+	kind tokKind
+	text string
+	val  float64
+}
+
+// lex scans s into a token stream terminated by tokEOF. Any ASCII letter run that
+// is not exactly "sqrt" is rejected.
+func lex(s string) ([]token, error) {
+	toks := make([]token, 0, len(s)/2+1)
+	for i := 0; i < len(s); {
+		c := s[i]
+		switch {
+		case c == ' ' || c == '\t' || c == '\n':
+			i++
+		case c == '(':
+			toks = append(toks, token{kind: tokLParen, text: "("})
+			i++
+		case c == ')':
+			toks = append(toks, token{kind: tokRParen, text: ")"})
+			i++
+		case c == '+' || c == '-' || c == '*' || c == '/' || c == '^':
+			toks = append(toks, token{kind: tokOp, text: string(c)})
+			i++
+		case isAsciiLetter(c):
+			start := i
+			for i < len(s) && isAsciiLetter(s[i]) {
+				i++
+			}
+			name := s[start:i]
+			if name != "sqrt" {
+				return nil, fmt.Errorf("%w: unexpected identifier %q", ErrEvaluation, name)
+			}
+			toks = append(toks, token{kind: tokIdent, text: name})
+		case c >= '0' && c <= '9' || c == '.':
+			start := i
+			dot := false
+			for i < len(s) {
+				d := s[i]
+				if d >= '0' && d <= '9' {
+					i++
+					continue
+				}
+				if d == '.' && !dot {
+					dot = true
+					i++
+					continue
+				}
+				break
+			}
+			raw := s[start:i]
+			val, err := strconv.ParseFloat(raw, 64)
+			if err != nil {
+				return nil, fmt.Errorf("%w: invalid number %q", ErrEvaluation, raw)
+			}
+			toks = append(toks, token{kind: tokNum, text: raw, val: val})
+		default:
+			return nil, fmt.Errorf("%w: disallowed character %q", ErrEvaluation, c)
+		}
+	}
+	toks = append(toks, token{kind: tokEOF, text: ""})
+	return toks, nil
+}
+
+// parser is a recursive-descent parser over the token stream. Grammar:
+//
+//	expr   := term (('+'|'-') term)*
+//	term   := unary (('*'|'/') unary)*
+//	unary  := ('+'|'-') unary | power
+//	power  := primary ('^' unary)?
+//
+// ^ binds tighter than unary minus (so -2^2 == -(2^2)) and is right-associative.
+type parser struct {
+	toks []token
+	pos  int
+}
+
+func (p *parser) cur() token { return p.toks[p.pos] }
+
+func (p *parser) parse() (float64, error) {
+	v, err := p.expr()
+	if err != nil {
+		return 0, err
+	}
+	if p.cur().kind != tokEOF {
+		return 0, fmt.Errorf("%w: unexpected token %q", ErrEvaluation, p.cur().text)
+	}
+	return v, nil
+}
+
+func (p *parser) expr() (float64, error) {
+	v, err := p.term()
+	if err != nil {
+		return 0, err
+	}
+	for p.cur().kind == tokOp && (p.cur().text == "+" || p.cur().text == "-") {
+		op := p.cur().text
+		p.pos++
+		r, err := p.term()
+		if err != nil {
+			return 0, err
+		}
+		if op == "+" {
+			v += r
+		} else {
+			v -= r
+		}
+	}
+	return v, nil
+}
+
+func (p *parser) term() (float64, error) {
+	v, err := p.unary()
+	if err != nil {
+		return 0, err
+	}
+	for p.cur().kind == tokOp && (p.cur().text == "*" || p.cur().text == "/") {
+		op := p.cur().text
+		p.pos++
+		r, err := p.unary()
+		if err != nil {
+			return 0, err
+		}
+		if op == "*" {
+			v *= r
+		} else {
+			if r == 0 {
+				return 0, fmt.Errorf("%w: division by zero", ErrEvaluation)
+			}
+			v /= r
+		}
+	}
+	return v, nil
+}
+
+func (p *parser) unary() (float64, error) {
+	if p.cur().kind == tokOp && (p.cur().text == "+" || p.cur().text == "-") {
+		neg := p.cur().text == "-"
+		p.pos++
+		v, err := p.unary()
+		if err != nil {
+			return 0, err
+		}
+		if neg {
+			return -v, nil
+		}
+		return v, nil
+	}
+	return p.power()
+}
+
+func (p *parser) power() (float64, error) {
+	base, err := p.primary()
+	if err != nil {
+		return 0, err
+	}
+	if p.cur().kind == tokOp && p.cur().text == "^" {
+		p.pos++
+		exp, err := p.unary()
+		if err != nil {
+			return 0, err
+		}
+		r := math.Pow(base, exp)
+		if math.IsNaN(r) {
+			return 0, fmt.Errorf("%w: non-real result", ErrEvaluation)
+		}
+		return r, nil
+	}
+	return base, nil
+}
+
+func (p *parser) primary() (float64, error) {
+	t := p.cur()
+	switch t.kind {
+	case tokNum:
+		p.pos++
+		return t.val, nil
+	case tokLParen:
+		p.pos++
+		v, err := p.expr()
+		if err != nil {
+			return 0, err
+		}
+		if p.cur().kind != tokRParen {
+			return 0, fmt.Errorf("%w: missing closing parenthesis", ErrEvaluation)
+		}
+		p.pos++
+		return v, nil
+	case tokIdent:
+		p.pos++
+		if p.cur().kind != tokLParen {
+			return 0, fmt.Errorf("%w: sqrt requires parentheses", ErrEvaluation)
+		}
+		p.pos++
+		v, err := p.expr()
+		if err != nil {
+			return 0, err
+		}
+		if p.cur().kind != tokRParen {
+			return 0, fmt.Errorf("%w: missing closing parenthesis in sqrt", ErrEvaluation)
+		}
+		p.pos++
+		if v < 0 {
+			return 0, fmt.Errorf("%w: sqrt of a negative number", ErrEvaluation)
+		}
+		return math.Sqrt(v), nil
+	default:
+		return 0, fmt.Errorf("%w: unexpected token %q", ErrEvaluation, t.text)
+	}
+}
diff --git a/plugins/tools/basic/calculator_test.go b/plugins/tools/basic/calculator_test.go
new file mode 100644
index 0000000..2f39d5f
--- /dev/null
+++ b/plugins/tools/basic/calculator_test.go
@@ -0,0 +1,113 @@
+package basic
+
+import (
+	"errors"
+	"strings"
+	"testing"
+)
+
+// TestCalculator_Evaluate_Normal covers the arithmetic grammar.
+func TestCalculator_Evaluate_Normal(t *testing.T) {
+	t.Parallel()
+	c := calculator{}
+	cases := []struct {
+		expr string
+		want float64
+	}{
+		{expr: "1+2*3", want: 7},
+		{expr: "(1+2)*3", want: 9},
+		{expr: "2^10", want: 1024},
+		{expr: "2^3^2", want: 512},  // right-associative
+		{expr: "-2^2", want: -4},    // power binds tighter than unary minus
+		{expr: "2^-3", want: 0.125}, // negative exponent
+		{expr: "sqrt(16)", want: 4},
+		{expr: "sqrt(2)+1", want: 2.414213562373095},
+		{expr: "10/4", want: 2.5},
+		{expr: "  ( 5 + 3 ) * 2 ", want: 16},
+	}
+	for _, tc := range cases {
+		tc := tc
+		t.Run(tc.expr, func(t *testing.T) {
+			t.Parallel()
+			got, err := run(t, c, map[string]any{"expression": tc.expr})
+			assertNoErr(t, err)
+			assertApprox(t, fieldFloat(t, mapResult(t, got), "result"), tc.want)
+		})
+	}
+}
+
+// TestCalculator_Evaluate_Errors covers invalid expressions and missing/typed args.
+func TestCalculator_Evaluate_Errors(t *testing.T) {
+	t.Parallel()
+	c := calculator{}
+	byExpression := []struct {
+		name string
+		args map[string]any
+	}{
+		{name: "missing expression", args: map[string]any{}},
+		{name: "non-string expression", args: map[string]any{"expression": 42}},
+		{name: "division by zero", args: map[string]any{"expression": "1/0"}},
+		{name: "trailing operator", args: map[string]any{"expression": "1+"}},
+		{name: "bare identifier", args: map[string]any{"expression": "x+1"}},
+		{name: "unknown function", args: map[string]any{"expression": "sqrt(16"}},
+		{name: "negative sqrt", args: map[string]any{"expression": "sqrt(-1)"}},
+		{name: "empty string", args: map[string]any{"expression": ""}},
+		{name: "non-ASCII injection", args: map[string]any{"expression": "1+2;DROP"}},
+	}
+	for _, tc := range byExpression {
+		tc := tc
+		t.Run(tc.name, func(t *testing.T) {
+			t.Parallel()
+			_, err := run(t, c, tc.args)
+			if err == nil {
+				t.Fatalf("want error, got nil")
+			}
+			if !errors.Is(err, ErrInvalidArgument) && !errors.Is(err, ErrEvaluation) {
+				t.Fatalf("err %v wraps neither ErrInvalidArgument nor ErrEvaluation", err)
+			}
+		})
+	}
+}
+
+// TestCalculator_InjectionRejected ensures expression injection is refused.
+func TestCalculator_InjectionRejected(t *testing.T) {
+	t.Parallel()
+	c := calculator{}
+	payloads := []string{
+		"os.system('id')",
+		"process.env.HOME",
+		`__import__("os")`,
+		"1+2*3;rm -rf /",
+		"`id`",
+		"eval(1)",
+		"1 && 2",
+		"len(open('x'))",
+	}
+	for _, p := range payloads {
+		p := p
+		t.Run(p, func(t *testing.T) {
+			t.Parallel()
+			_, err := run(t, c, map[string]any{"expression": p})
+			assertIs(t, err, ErrEvaluation)
+		})
+	}
+	// The expression length bound rejects a long payload.
+	large := strings.Repeat("1+", 200) + "1"
+	_, err := run(t, c, map[string]any{"expression": large})
+	assertIs(t, err, ErrEvaluation)
+}
+
+// TestCalculator_Schema verifies the JSON Schema contract.
+func TestCalculator_Schema(t *testing.T) {
+	t.Parallel()
+	p, err := calculator{}.Parameters()
+	assertNoErr(t, err)
+	reqs := p["required"].([]any)
+	if len(reqs) != 1 || reqs[0] != "expression" {
+		t.Fatalf("required = %v, want [expression]", reqs)
+	}
+	props := p["properties"].(map[string]any)
+	if got := props["expression"].(map[string]any)["type"]; got != "string" {
+		t.Fatalf("expression type = %v, want string", got)
+	}
+}
diff --git a/plugins/tools/basic/context_test.go b/plugins/tools/basic/context_test.go
new file mode 100644
index 0000000..bf9fd13
--- /dev/null
+++ b/plugins/tools/basic/context_test.go
@@ -0,0 +1,25 @@
+package basic
+
+import (
+	"context"
+	"errors"
+	"testing"
+)
+
+// TestExecute_HonorsCancellation verifies every tool returns the context error
+// when ctx is already canceled, without performing any work.
+func TestExecute_HonorsCancellation(t *testing.T) {
+	t.Parallel()
+	ctx, cancel := context.WithCancel(context.Background())
+	cancel()
+	for _, tool := range Tools() {
+		tool := tool
+		t.Run(tool.Name(), func(t *testing.T) {
+			t.Parallel()
+			_, err := tool.Execute(ctx, map[string]any{})
+			if !errors.Is(err, context.Canceled) {
+				t.Fatalf("err = %v, want wrapping context.Canceled", err)
+			}
+		})
+	}
+}
diff --git a/plugins/tools/basic/datetime.go b/plugins/tools/basic/datetime.go
new file mode 100644
index 0000000..4b641b2
--- /dev/null
+++ b/plugins/tools/basic/datetime.go
@@ -0,0 +1,77 @@
+package basic
+
+import (
+	"context"
+	"fmt"
+	"time"
+)
+
+// datetime returns the current time, optionally formatted with a Go reference
+// layout and rendered in a named IANA timezone.
+type datetime struct{}
+
+const (
+	// defaultDTFormat is the layout used when the caller supplies none.
+	defaultDTFormat = time.RFC3339
+	// defaultDTZone is the timezone used when the caller supplies none.
+	defaultDTZone = "UTC"
+	// maxDTFormatLen bounds a layout string to guard against absurd input.
+	maxDTFormatLen = 128
+	// maxDTZoneLen bounds a timezone name to guard against absurd input.
+	maxDTZoneLen = 64
+)
+
+// Name returns the tool identifier.
+func (datetime) Name() string { return "datetime" }
+
+// Description explains what the tool does for the model.
+func (datetime) Description() string {
+	return "Return the current time, formatted with a Go reference layout and rendered in a named IANA timezone."
+}
+
+// Parameters returns the tool input JSON Schema.
+func (datetime) Parameters() (map[string]any, error) {
+	return map[string]any{
+		"type": "object",
+		"properties": map[string]any{
+			"format":   map[string]any{"type": "string", "description": "Go reference layout, e.g. 2006-01-02 15:04:05"},
+			"timezone": map[string]any{"type": "string", "description": "IANA timezone name, e.g. Asia/Shanghai"},
+		},
+	}, nil
+}
+
+// Execute returns the current time in the requested format and timezone.
+func (datetime) Execute(ctx context.Context, args any) (any, error) {
+	if err := ctx.Err(); err != nil {
+		return nil, err
+	}
+	m, err := asMap(args)
+	if err != nil {
+		return nil, err
+	}
+	layout, err := optString(m, "format", defaultDTFormat)
+	if err != nil {
+		return nil, err
+	}
+	if len(layout) > maxDTFormatLen {
+		return nil, fmt.Errorf("basic: layout too long: %w", ErrInvalidArgument)
+	}
+	zone, err := optString(m, "timezone", defaultDTZone)
+	if err != nil {
+		return nil, err
+	}
+	if len(zone) > maxDTZoneLen {
+		return nil, fmt.Errorf("basic: timezone name too long: %w", ErrInvalidArgument)
+	}
+	loc, err := time.LoadLocation(zone)
+	if err != nil {
+		return nil, fmt.Errorf("basic: invalid timezone %q: %w", zone, err)
+	}
+	now := time.Now().In(loc)
+	return map[string]any{
+		"now":      now.Format(layout),
+		"unix":     now.Unix(),
+		"timezone": zone,
+		"format":   layout,
+	}, nil
+}
diff --git a/plugins/tools/basic/datetime_test.go b/plugins/tools/basic/datetime_test.go
new file mode 100644
index 0000000..a93b0d8
--- /dev/null
+++ b/plugins/tools/basic/datetime_test.go
@@ -0,0 +1,99 @@
+package basic
+
+import (
+	"testing"
+	"time"
+)
+
+// TestDatetime_Execute_Normal covers default, formatted and zone-shifted output.
+func TestDatetime_Execute_Normal(t *testing.T) {
+	t.Parallel()
+	c := datetime{}
+
+	t.Run("default", func(t *testing.T) {
+		t.Parallel()
+		got, err := run(t, c, nil)
+		assertNoErr(t, err)
+		m := mapResult(t, got)
+		if fieldString(t, m, "timezone") != "UTC" {
+			t.Fatalf("timezone = %v, want UTC", m["timezone"])
+		}
+		if fieldString(t, m, "format") != time.RFC3339 {
+			t.Fatalf("format = %v, want RFC3339", m["format"])
+		}
+		if _, err := time.Parse(time.RFC3339, fieldString(t, m, "now")); err != nil {
+			t.Fatalf("now %q is not RFC3339: %v", m["now"], err)
+		}
+		if fieldInt64(t, m, "unix") <= 0 {
+			t.Fatalf("unix = %v, want positive", m["unix"])
+		}
+	})
+
+	t.Run("custom format", func(t *testing.T) {
+		t.Parallel()
+		got, err := run(t, c, map[string]any{"format": "2006-01-02"})
+		assertNoErr(t, err)
+		m := mapResult(t, got)
+		if len(fieldString(t, m, "now")) != 10 {
+			t.Fatalf("now = %v, want 10-char date", m["now"])
+		}
+	})
+
+	t.Run("non-UTC zone", func(t *testing.T) {
+		t.Parallel()
+		if _, err := time.LoadLocation("Asia/Shanghai"); err != nil {
+			t.Skipf("tzdata not available: %v", err)
+		}
+		got, err := run(t, c, map[string]any{"timezone": "Asia/Shanghai"})
+		assertNoErr(t, err)
+		m := mapResult(t, got)
+		if fieldString(t, m, "timezone") != "Asia/Shanghai" {
+			t.Fatalf("timezone = %v, want Asia/Shanghai", m["timezone"])
+		}
+	})
+}
+
+// TestDatetime_Execute_Errors covers invalid zones and overlong inputs.
+func TestDatetime_Execute_Errors(t *testing.T) {
+	t.Parallel()
+	c := datetime{}
+	cases := []struct {
+		name   string
+		args   map[string]any
+		target error
+	}{
+		{name: "invalid zone", args: map[string]any{"timezone": "Not/AZone"}, target: nil},
+		{name: "zone too long", args: map[string]any{"timezone": string(make([]byte, 65))}, target: ErrInvalidArgument},
+		{name: "layout too long", args: map[string]any{"format": string(make([]byte, 129))}, target: ErrInvalidArgument},
+		{name: "non-string timezone", args: map[string]any{"timezone": 123}, target: ErrInvalidArgument},
+	}
+	for _, tc := range cases {
+		tc := tc
+		t.Run(tc.name, func(t *testing.T) {
+			t.Parallel()
+			_, err := run(t, c, tc.args)
+			if err == nil {
+				t.Fatalf("want error, got nil")
+			}
+			if tc.target != nil {
+				assertIs(t, err, tc.target)
+			}
+		})
+	}
+}
+
+// TestDatetime_Schema verifies the JSON Schema contract.
+func TestDatetime_Schema(t *testing.T) {
+	t.Parallel()
+	p, err := datetime{}.Parameters()
+	assertNoErr(t, err)
+	if _, ok := p["required"].([]any); ok {
+		t.Fatalf("datetime schema should have no required fields")
+	}
+	props := p["properties"].(map[string]any)
+	for _, field := range []string{"format", "timezone"} {
+		if props[field].(map[string]any)["type"] != "string" {
+			t.Fatalf("%s type should be string", field)
+		}
+	}
+}
diff --git a/plugins/tools/basic/doc.go b/plugins/tools/basic/doc.go
new file mode 100644
index 0000000..c8ac610
--- /dev/null
+++ b/plugins/tools/basic/doc.go
@@ -0,0 +1,9 @@
+// Package basic implements the built-in foundation tools for Cyrene v2
+// (DevDocs/design/09-插件生态.md, DevDocs/开发时间线.md M5): calculator,
+// datetime, random, text, json and markdown.
+//
+// Each tool is an sdk.Tool that is registered through the capability registry
+// via Register, and all of them are stateless values, so they are safe for
+// concurrent use. The tools are pure — no I/O and no network — so they can be
+// invoked from any role, any mode and any channel without sandboxing.
+package basic
diff --git a/plugins/tools/basic/helpers_test.go b/plugins/tools/basic/helpers_test.go
new file mode 100644
index 0000000..eec3b5a
--- /dev/null
+++ b/plugins/tools/basic/helpers_test.go
@@ -0,0 +1,104 @@
+package basic
+
+import (
+	"context"
+	"errors"
+	"math"
+	"testing"
+)
+
+// executor is the subset of sdk.Tool used in tests so concrete tools can be
+// exercised without importing the SDK into the test table.
+type executor interface {
+	Execute(context.Context, any) (any, error)
+}
+
+// run executes e with args against a background context.
+func run(t *testing.T, e executor, args map[string]any) (any, error) {
+	t.Helper()
+	return e.Execute(context.Background(), args)
+}
+
+// assertIs fails when err does not wrap target.
+func assertIs(t *testing.T, err, target error) {
+	t.Helper()
+	if !errors.Is(err, target) {
+		t.Fatalf("errors.Is(%v, %v) = false, want true", err, target)
+	}
+}
+
+// assertNoErr fails when err is non-nil.
+func assertNoErr(t *testing.T, err error) {
+	t.Helper()
+	if err != nil {
+		t.Fatalf("unexpected error: %v", err)
+	}
+}
+
+// mapResult asserts got is a map[string]any and returns it.
+func mapResult(t *testing.T, got any) map[string]any {
+	t.Helper()
+	m, ok := got.(map[string]any)
+	if !ok {
+		t.Fatalf("result is %T, want map[string]any", got)
+	}
+	return m
+}
+
+// fieldString returns m[key] as a string, failing otherwise.
+func fieldString(t *testing.T, m map[string]any, key string) string {
+	t.Helper()
+	v, ok := m[key].(string)
+	if !ok {
+		t.Fatalf("%s field is %T (%v), want string", key, m[key], m[key])
+	}
+	return v
+}
+
+// fieldFloat returns m[key] as a float64, failing otherwise.
+func fieldFloat(t *testing.T, m map[string]any, key string) float64 {
+	t.Helper()
+	v, ok := m[key].(float64)
+	if !ok {
+		t.Fatalf("%s field is %T (%v), want float64", key, m[key], m[key])
+	}
+	return v
+}
+
+// fieldInt returns m[key] as an int, failing otherwise.
+func fieldInt(t *testing.T, m map[string]any, key string) int {
+	t.Helper()
+	v, ok := m[key].(int)
+	if !ok {
+		t.Fatalf("%s field is %T (%v), want int", key, m[key], m[key])
+	}
+	return v
+}
+
+// fieldInt64 returns m[key] as an int64, failing otherwise.
+func fieldInt64(t *testing.T, m map[string]any, key string) int64 {
+	t.Helper()
+	v, ok := m[key].(int64)
+	if !ok {
+		t.Fatalf("%s field is %T (%v), want int64", key, m[key], m[key])
+	}
+	return v
+}
+
+// fieldBool returns m[key] as a bool, failing otherwise.
+func fieldBool(t *testing.T, m map[string]any, key string) bool {
+	t.Helper()
+	v, ok := m[key].(bool)
+	if !ok {
+		t.Fatalf("%s field is %T (%v), want bool", key, m[key], m[key])
+	}
+	return v
+}
+
+// assertApprox fails when got and want differ by more than 1e-9.
+func assertApprox(t *testing.T, got, want float64) {
+	t.Helper()
+	if math.Abs(got-want) > 1e-9 {
+		t.Fatalf("got %v, want %v", got, want)
+	}
+}
diff --git a/plugins/tools/basic/json.go b/plugins/tools/basic/json.go
new file mode 100644
index 0000000..b82f006
--- /dev/null
+++ b/plugins/tools/basic/json.go
@@ -0,0 +1,95 @@
+package basic
+
+import (
+	"bytes"
+	"context"
+	"encoding/json"
+	"fmt"
+	"strings"
+)
+
+// jsonTool exposes JSON parsing, formatting, validation and serialization for the
+// agent. `value` is unconstrained in the schema because it is a string for
+// parse/format/validate and any JSON value for stringify.
+type jsonTool struct{}
+
+// Name returns the tool identifier.
+func (jsonTool) Name() string { return "json" }
+
+// Description explains what the tool does for the model.
+func (jsonTool) Description() string {
+	return "Parse, format, validate or serialize JSON for structured data handling."
+}
+
+// Parameters returns the tool input JSON Schema.
+func (jsonTool) Parameters() (map[string]any, error) {
+	return map[string]any{
+		"type": "object",
+		"properties": map[string]any{
+			"operation": map[string]any{
+				"type":        "string",
+				"enum":        []any{"parse", "format", "validate", "stringify"},
+				"description": "the operation to perform",
+			},
+			"value": map[string]any{"description": "input: a JSON string for parse/format/validate, any value for stringify"},
+		},
+		"required": []any{"operation"},
+	}, nil
+}
+
+// Execute runs the requested operation against value.
+func (jsonTool) Execute(ctx context.Context, args any) (any, error) {
+	if err := ctx.Err(); err != nil {
+		return nil, err
+	}
+	m, err := asMap(args)
+	if err != nil {
+		return nil, err
+	}
+	op, err := reqString(m, "operation")
+	if err != nil {
+		return nil, err
+	}
+	if !hasKey(m, "value") {
+		return nil, fmt.Errorf("basic: missing required argument %q: %w", "value", ErrInvalidArgument)
+	}
+	value := m["value"]
+	switch strings.ToLower(strings.TrimSpace(op)) {
+	case "parse":
+		s, ok := value.(string)
+		if !ok {
+			return nil, fmt.Errorf("basic: argument %q must be a string for parse: %w", "value", ErrInvalidArgument)
+		}
+		var out any
+		if err := json.Unmarshal([]byte(s), &out); err != nil {
+			return nil, fmt.Errorf("basic: invalid json for parse: %w", err)
+		}
+		return map[string]any{"result": out}, nil
+	case "format":
+		s, ok := value.(string)
+		if !ok {
+			return nil, fmt.Errorf("basic: argument %q must be a string for format: %w", "value", ErrInvalidArgument)
+		}
+		var buf bytes.Buffer
+		if err := json.Indent(&buf, []byte(s), "", "  "); err != nil {
+			return nil, fmt.Errorf("basic: invalid json for format: %w", err)
+		}
+		return map[string]any{"result": buf.String()}, nil
+	case "validate":
+		s, ok := value.(string)
+		if !ok {
+			return nil, fmt.Errorf("basic: argument %q must be a string for validate: %w", "value", ErrInvalidArgument)
+		}
+		return map[string]any{"valid": json.Valid([]byte(s))}, nil
+	case "stringify":
+		var buf bytes.Buffer
+		enc := json.NewEncoder(&buf)
+		enc.SetIndent("", "  ")
+		if err := enc.Encode(value); err != nil {
+			return nil, fmt.Errorf("basic: cannot serialize value: %w", err)
+		}
+		return map[string]any{"result": strings.TrimRight(buf.String(), "\n")}, nil
+	default:
+		return nil, fmt.Errorf("basic: unknown json operation %q: %w", op, ErrUnsupported)
+	}
+}
diff --git a/plugins/tools/basic/json_test.go b/plugins/tools/basic/json_test.go
new file mode 100644
index 0000000..ef021a7
--- /dev/null
+++ b/plugins/tools/basic/json_test.go
@@ -0,0 +1,119 @@
+package basic
+
+import (
+	"reflect"
+	"testing"
+)
+
+// TestJson_Execute_Normal covers parse, format, validate and stringify.
+func TestJson_Execute_Normal(t *testing.T) {
+	t.Parallel()
+	tool := jsonTool{}
+
+	t.Run("parse", func(t *testing.T) {
+		t.Parallel()
+		got, err := run(t, tool, map[string]any{"operation": "parse", "value": `{"a":1,"b":[true,null,"s"]}`})
+		assertNoErr(t, err)
+		m := mapResult(t, got)
+		obj := m["result"].(map[string]any)
+		if obj["a"] != float64(1) {
+			t.Fatalf("a = %v, want 1", obj["a"])
+		}
+		if !reflect.DeepEqual(obj["b"], []any{true, nil, "s"}) {
+			t.Fatalf("b = %v, want [true nil s]", obj["b"])
+		}
+	})
+
+	t.Run("string value parse", func(t *testing.T) {
+		t.Parallel()
+		got, err := run(t, tool, map[string]any{"operation": "parse", "value": `"hi"`})
+		assertNoErr(t, err)
+		if fieldString(t, mapResult(t, got), "result") != "hi" {
+			t.Fatalf("result should be hi")
+		}
+	})
+
+	t.Run("format", func(t *testing.T) {
+		t.Parallel()
+		got, err := run(t, tool, map[string]any{"operation": "format", "value": `{"x":1}`})
+		assertNoErr(t, err)
+		m := mapResult(t, got)
+		if want := "{\n  \"x\": 1\n}"; fieldString(t, m, "result") != want {
+			t.Fatalf("result = %q, want %q", m["result"], want)
+		}
+	})
+
+	t.Run("validate valid", func(t *testing.T) {
+		t.Parallel()
+		got, err := run(t, tool, map[string]any{"operation": "validate", "value": `{"a":[1,2]}`})
+		assertNoErr(t, err)
+		if !fieldBool(t, mapResult(t, got), "valid") {
+			t.Fatalf("valid should be true")
+		}
+	})
+
+	t.Run("validate invalid", func(t *testing.T) {
+		t.Parallel()
+		got, err := run(t, tool, map[string]any{"operation": "validate", "value": `{broken`})
+		assertNoErr(t, err)
+		if fieldBool(t, mapResult(t, got), "valid") {
+			t.Fatalf("valid should be false")
+		}
+	})
+
+	t.Run("stringify", func(t *testing.T) {
+		t.Parallel()
+		got, err := run(t, tool, map[string]any{"operation": "stringify", "value": map[string]any{"x": 1}})
+		assertNoErr(t, err)
+		m := mapResult(t, got)
+		if want := "{\n  \"x\": 1\n}"; fieldString(t, m, "result") != want {
+			t.Fatalf("result = %q, want %q", m["result"], want)
+		}
+	})
+}
+
+// TestJson_Execute_Errors covers invalid inputs and unsupported operations.
+func TestJson_Execute_Errors(t *testing.T) {
+	t.Parallel()
+	tool := jsonTool{}
+	cases := []struct {
+		name   string
+		args   map[string]any
+		target error
+	}{
+		{name: "invalid parse", args: map[string]any{"operation": "parse", "value": `{bad`}, target: nil},
+		{name: "parse non-string", args: map[string]any{"operation": "parse", "value": 42}, target: ErrInvalidArgument},
+		{name: "invalid format", args: map[string]any{"operation": "format", "value": `{bad`}, target: nil},
+		{name: "validate non-string", args: map[string]any{"operation": "validate", "value": true}, target: ErrInvalidArgument},
+		{name: "unknown operation", args: map[string]any{"operation": "csv", "value": "{}"}, target: ErrUnsupported},
+		{name: "missing value", args: map[string]any{"operation": "stringify"}, target: ErrInvalidArgument},
+		{name: "missing operation", args: map[string]any{}, target: ErrInvalidArgument},
+	}
+	for _, tc := range cases {
+		tc := tc
+		t.Run(tc.name, func(t *testing.T) {
+			t.Parallel()
+			_, err := run(t, tool, tc.args)
+			if err == nil {
+				t.Fatalf("want error, got nil")
+			}
+			if tc.target != nil {
+				assertIs(t, err, tc.target)
+			}
+		})
+	}
+}
+
+// TestJson_Schema verifies required operation.
+func TestJson_Schema(t *testing.T) {
+	t.Parallel()
+	p, err := jsonTool{}.Parameters()
+	assertNoErr(t, err)
+	reqs := p["required"].([]any)
+	if len(reqs) != 1 || reqs[0] != "operation" {
+		t.Fatalf("required = %v, want [operation]", reqs)
+	}
+	if _, ok := p["properties"].(map[string]any)["value"]; !ok {
+		t.Fatalf("schema should include a value property")
+	}
+}
diff --git a/plugins/tools/basic/log.go b/plugins/tools/basic/log.go
new file mode 100644
index 0000000..74f0f77
--- /dev/null
+++ b/plugins/tools/basic/log.go
@@ -0,0 +1,17 @@
+package basic
+
+import (
+	"log/slog"
+
+	"git.yeij.top/AskaEth/Cyrene_v2/internal/logger"
+)
+
+// log returns a logger tagged with this plugin's source so every record it emits
+// is traceable back to its owning module (DevDocs/02-development-constraints.md
+// §五). The sdk.Tool surface does not carry an injected base logger, so a fresh
+// one is built on demand; it is only used on the error path, where the cost is
+// negligible. Errors are also returned to the caller so the failure is never
+// swallowed.
+func log() *slog.Logger {
+	return logger.WithSource(logger.New(logger.Options{}), "plugins/tools/basic")
+}
diff --git a/plugins/tools/basic/markdown.go b/plugins/tools/basic/markdown.go
new file mode 100644
index 0000000..1fe9735
--- /dev/null
+++ b/plugins/tools/basic/markdown.go
@@ -0,0 +1,193 @@
+package basic
+
+import (
+	"context"
+	"fmt"
+	"strings"
+)
+
+// markdown exposes Markdown building and parsing helpers: tables, lists, headings
+// and YAML front-matter metadata.
+type markdown struct{}
+
+const (
+	// minHeadingLevel and maxHeadingLevel bound the supported ATX heading depth.
+	minHeadingLevel = 1
+	maxHeadingLevel = 6
+)
+
+// Name returns the tool identifier.
+func (markdown) Name() string { return "markdown" }
+
+// Description explains what the tool does for the model.
+func (markdown) Description() string {
+	return "Build or parse Markdown: tables, lists, headings and YAML front-matter metadata."
+}
+
+// Parameters returns the tool input JSON Schema. Operation-specific inputs are
+// validated by Execute.
+func (markdown) Parameters() (map[string]any, error) {
+	return map[string]any{
+		"type": "object",
+		"properties": map[string]any{
+			"operation": map[string]any{
+				"type":        "string",
+				"enum":        []any{"table", "list", "heading", "metadata"},
+				"description": "the operation to perform",
+			},
+			"content": map[string]any{"type": "string", "description": "Markdown body for metadata parsing"},
+			"headers": map[string]any{"type": "array", "description": "table column headers"},
+			"rows":    map[string]any{"type": "array", "description": "table rows (array of arrays of strings)"},
+			"items":   map[string]any{"type": "array", "description": "list entries"},
+			"ordered": map[string]any{"type": "boolean", "description": "render a numbered list when true"},
+			"level":   map[string]any{"type": "integer", "description": "heading level 1..6"},
+			"text":    map[string]any{"type": "string", "description": "heading text"},
+		},
+		"required": []any{"operation"},
+	}, nil
+}
+
+// Execute runs the requested operation.
+func (markdown) Execute(ctx context.Context, args any) (any, error) {
+	if err := ctx.Err(); err != nil {
+		return nil, err
+	}
+	m, err := asMap(args)
+	if err != nil {
+		return nil, err
+	}
+	op, err := reqString(m, "operation")
+	if err != nil {
+		return nil, err
+	}
+	switch strings.ToLower(strings.TrimSpace(op)) {
+	case "table":
+		return mdTable(m)
+	case "list":
+		return mdList(m)
+	case "heading":
+		return mdHeading(m)
+	case "metadata":
+		return mdMeta(m)
+	default:
+		return nil, fmt.Errorf("basic: unknown markdown operation %q: %w", op, ErrUnsupported)
+	}
+}
+
+// mdTable builds a Markdown table from headers and rows.
+func mdTable(m map[string]any) (any, error) {
+	headers, err := strSliceArg(m, "headers")
+	if err != nil {
+		return nil, err
+	}
+	rows, err := strSliceSliceArg(m, "rows")
+	if err != nil {
+		return nil, err
+	}
+	var b strings.Builder
+	write := func(cells []string) {
+		b.WriteString("| ")
+		for i, c := range cells {
+			if i > 0 {
+				b.WriteString(" | ")
+			}
+			b.WriteString(escapeCell(c))
+		}
+		b.WriteString(" |\n")
+	}
+	write(headers)
+	b.WriteString(tableSeparator(headers))
+	for _, r := range rows {
+		write(r)
+	}
+	return map[string]any{"result": b.String()}, nil
+}
+
+// tableSeparator returns the | --- | --- | header separator row for the column count.
+func tableSeparator(cols []string) string {
+	parts := make([]string, len(cols))
+	for i := range parts {
+		parts[i] = "---"
+	}
+	return "| " + strings.Join(parts, " | ") + " |\n"
+}
+
+// escapeCell neutralizes a cell so a pipe inside the value does not break the table.
+func escapeCell(s string) string {
+	s = strings.ReplaceAll(s, "|", `\|`)
+	s = strings.ReplaceAll(s, "\n", " ")
+	return strings.ReplaceAll(s, "\r", "")
+}
+
+// mdList builds a bullet or numbered list from items.
+func mdList(m map[string]any) (any, error) {
+	items, err := strSliceArg(m, "items")
+	if err != nil {
+		return nil, err
+	}
+	ordered, err := optBool(m, "ordered", false)
+	if err != nil {
+		return nil, err
+	}
+	var b strings.Builder
+	for i, it := range items {
+		if ordered {
+			fmt.Fprintf(&b, "%d. %s\n", i+1, it)
+		} else {
+			b.WriteString("- " + it + "\n")
+		}
+	}
+	return map[string]any{"result": b.String()}, nil
+}
+
+// mdHeading builds a Markdown ATX heading.
+func mdHeading(m map[string]any) (any, error) {
+	level, err := optInt(m, "level", 1)
+	if err != nil {
+		return nil, err
+	}
+	text, err := reqString(m, "text")
+	if err != nil {
+		return nil, err
+	}
+	if level < minHeadingLevel || level > maxHeadingLevel {
+		return nil, fmt.Errorf("basic: heading level must be between %d and %d: %w", minHeadingLevel, maxHeadingLevel, ErrInvalidArgument)
+	}
+	return map[string]any{"result": strings.Repeat("#", level) + " " + text + "\n"}, nil
+}
+
+// mdMeta parses YAML front matter from content and returns the metadata table plus
+// the remaining body.
+func mdMeta(m map[string]any) (any, error) {
+	content, err := reqString(m, "content")
+	if err != nil {
+		return nil, err
+	}
+	meta, body := parseFrontMatter(content)
+	return map[string]any{"meta": meta, "body": body}, nil
+}
+
+// parseFrontMatter splits leading `--- ... ---` front matter from a Markdown body.
+// When the document has no front matter it returns an empty table and the whole
+// content as the body.
+func parseFrontMatter(content string) (map[string]string, string) {
+	lines := strings.Split(content, "\n")
+	if len(lines) == 0 || strings.TrimSpace(lines[0]) != "---" {
+		return map[string]string{}, content
+	}
+	meta := map[string]string{}
+	i := 1
+	for i < len(lines) {
+		line := strings.TrimRight(lines[i], "\r")
+		if strings.TrimSpace(line) == "---" {
+			i++
+			break
+		}
+		if key, val, ok := strings.Cut(line, ":"); ok {
+			meta[strings.TrimSpace(key)] = strings.TrimSpace(val)
+		}
+		i++
+	}
+	body := strings.Join(lines[i:], "\n")
+	return meta, body
+}
diff --git a/plugins/tools/basic/markdown_test.go b/plugins/tools/basic/markdown_test.go
new file mode 100644
index 0000000..33258d6
--- /dev/null
+++ b/plugins/tools/basic/markdown_test.go
@@ -0,0 +1,121 @@
+package basic
+
+import (
+	"strings"
+	"testing"
+)
+
+// TestMarkdown_Execute_Normal covers table, list, heading and metadata ops.
+func TestMarkdown_Execute_Normal(t *testing.T) {
+	t.Parallel()
+	tool := markdown{}
+
+	t.Run("table", func(t *testing.T) {
+		t.Parallel()
+		got, err := run(t, tool, map[string]any{
+			"operation": "table",
+			"headers":   []any{"name", "score"},
+			"rows":      []any{[]any{"ada", "90"}, []any{"grace", "100"}},
+		})
+		assertNoErr(t, err)
+		want := "| name | score |\n| --- | --- |\n| ada | 90 |\n| grace | 100 |\n"
+		if got := fieldString(t, mapResult(t, got), "result"); got != want {
+			t.Fatalf("table = %q, want %q", got, want)
+		}
+	})
+
+	t.Run("bulleted list", func(t *testing.T) {
+		t.Parallel()
+		got, err := run(t, tool, map[string]any{"operation": "list", "items": []any{"a", "b"}})
+		assertNoErr(t, err)
+		if want := "- a\n- b\n"; fieldString(t, mapResult(t, got), "result") != want {
+			t.Fatalf("got %q", fieldString(t, mapResult(t, got), "result"))
+		}
+	})
+
+	t.Run("ordered list", func(t *testing.T) {
+		t.Parallel()
+		got, err := run(t, tool, map[string]any{"operation": "list", "items": []any{"a", "b"}, "ordered": true})
+		assertNoErr(t, err)
+		if want := "1. a\n2. b\n"; fieldString(t, mapResult(t, got), "result") != want {
+			t.Fatalf("got %q", fieldString(t, mapResult(t, got), "result"))
+		}
+	})
+
+	t.Run("heading", func(t *testing.T) {
+		t.Parallel()
+		got, err := run(t, tool, map[string]any{"operation": "heading", "level": 2, "text": "Title"})
+		assertNoErr(t, err)
+		if want := "## Title\n"; fieldString(t, mapResult(t, got), "result") != want {
+			t.Fatalf("got %q", fieldString(t, mapResult(t, got), "result"))
+		}
+	})
+
+	t.Run("metadata with front matter", func(t *testing.T) {
+		t.Parallel()
+		got, err := run(t, tool, map[string]any{"operation": "metadata", "content": "---\ntitle: Hi\n---\nBody text"})
+		assertNoErr(t, err)
+		m := mapResult(t, got)
+		meta := m["meta"].(map[string]string)
+		if meta["title"] != "Hi" {
+			t.Fatalf("meta = %v, want title=Hi", meta)
+		}
+		if got := m["body"].(string); got != "Body text" {
+			t.Fatalf("body = %q, want Body text", got)
+		}
+	})
+
+	t.Run("metadata without front matter", func(t *testing.T) {
+		t.Parallel()
+		got, err := run(t, tool, map[string]any{"operation": "metadata", "content": "just body"})
+		assertNoErr(t, err)
+		m := mapResult(t, got)
+		if meta := m["meta"].(map[string]string); len(meta) != 0 {
+			t.Fatalf("meta = %v, want empty", meta)
+		}
+		if got := m["body"].(string); got != "just body" {
+			t.Fatalf("body = %q", got)
+		}
+	})
+}
+
+// TestMarkdown_Execute_Errors covers bad operation inputs.
+func TestMarkdown_Execute_Errors(t *testing.T) {
+	t.Parallel()
+	tool := markdown{}
+	cases := []struct {
+		name   string
+		args   map[string]any
+		target error
+	}{
+		{name: "unknown operation", args: map[string]any{"operation": "quote"}, target: ErrUnsupported},
+		{name: "table without headers", args: map[string]any{"operation": "table", "rows": []any{}}, target: ErrInvalidArgument},
+		{name: "list without items", args: map[string]any{"operation": "list"}, target: ErrInvalidArgument},
+		{name: "heading level too low", args: map[string]any{"operation": "heading", "level": 0, "text": "x"}, target: ErrInvalidArgument},
+		{name: "heading level too high", args: map[string]any{"operation": "heading", "level": 7, "text": "x"}, target: ErrInvalidArgument},
+		{name: "heading without text", args: map[string]any{"operation": "heading", "level": 1}, target: ErrInvalidArgument},
+		{name: "metadata without content", args: map[string]any{"operation": "metadata"}, target: ErrInvalidArgument},
+	}
+	for _, tc := range cases {
+		tc := tc
+		t.Run(tc.name, func(t *testing.T) {
+			t.Parallel()
+			_, err := run(t, tool, tc.args)
+			assertIs(t, err, tc.target)
+		})
+	}
+}
+
+// TestMarkdown_Schema verifies required operation.
+func TestMarkdown_Schema(t *testing.T) {
+	t.Parallel()
+	p, err := markdown{}.Parameters()
+	assertNoErr(t, err)
+	reqs := p["required"].([]any)
+	if len(reqs) != 1 || reqs[0] != "operation" {
+		t.Fatalf("required = %v, want [operation]", reqs)
+	}
+	if !strings.Contains(p["properties"].(map[string]any)["operation"].(map[string]any)["enum"].([]any)[0].(string), "table") {
+		t.Fatalf("enum should include table")
+	}
+}
diff --git a/plugins/tools/basic/random.go b/plugins/tools/basic/random.go
new file mode 100644
index 0000000..52f2bfa
--- /dev/null
+++ b/plugins/tools/basic/random.go
@@ -0,0 +1,125 @@
+package basic
+
+import (
+	"context"
+	"crypto/rand"
+	"fmt"
+	"math/big"
+	"strings"
+)
+
+// random generates a cryptographically-random integer in [min,max] or a random
+// string of a given length drawn from a charset. Providing min/max selects the
+// integer mode; providing length selects the string mode.
+type random struct{}
+
+const (
+	// maxRandomLen bounds a generated string length.
+	maxRandomLen = 1024
+	// defaultCharset is the alphabet used when none is supplied.
+	defaultCharset = "abcdefghijklmnopqrstuvwxyzABCDEFGHIJKLMNOPQRSTUVWXYZ0123456789"
+)
+
+// Name returns the tool identifier.
+func (random) Name() string { return "random" }
+
+// Description explains what the tool does for the model.
+func (random) Description() string {
+	return "Generate a cryptographically-random integer in [min,max] or a random string of a given length."
+}
+
+// Parameters returns the tool input JSON Schema.
+func (random) Parameters() (map[string]any, error) {
+	return map[string]any{
+		"type": "object",
+		"properties": map[string]any{
+			"min":     map[string]any{"type": "integer", "description": "inclusive lower bound for an integer"},
+			"max":     map[string]any{"type": "integer", "description": "inclusive upper bound for an integer"},
+			"length":  map[string]any{"type": "integer", "description": "length of the random string"},
+			"charset": map[string]any{"type": "string", "description": "characters the string is drawn from"},
+		},
+	}, nil
+}
+
+// Execute produces either an integer or a string depending on which arguments
+// are supplied.
+func (random) Execute(ctx context.Context, args any) (any, error) {
+	if err := ctx.Err(); err != nil {
+		return nil, err
+	}
+	m, err := asMap(args)
+	if err != nil {
+		return nil, err
+	}
+	if hasKey(m, "min") || hasKey(m, "max") {
+		min, err := optInt(m, "min", 0)
+		if err != nil {
+			return nil, err
+		}
+		max, err := optInt(m, "max", 0)
+		if err != nil {
+			return nil, err
+		}
+		v, err := randInt(min, max)
+		if err != nil {
+			return nil, err
+		}
+		return map[string]any{"value": v}, nil
+	}
+	if hasKey(m, "length") {
+		n, err := reqInt(m, "length")
+		if err != nil {
+			return nil, err
+		}
+		charset, err := optString(m, "charset", defaultCharset)
+		if err != nil {
+			return nil, err
+		}
+		s, err := randString(n, charset)
+		if err != nil {
+			return nil, err
+		}
+		return map[string]any{"value": s}, nil
+	}
+	return nil, fmt.Errorf("basic: provide min/max for an integer or length for a string: %w", ErrInvalidArgument)
+}
+
+// randInt returns a uniformly-distributed integer in [min,max] inclusive, using
+// crypto/rand so the result is not predictable from the process state.
+func randInt(min, max int) (int, error) {
+	if max < min {
+		return 0, fmt.Errorf("basic: max must be >= min: %w", ErrInvalidArgument)
+	}
+	if min == max {
+		return min, nil
+	}
+	span := int64(max) - int64(min) + 1
+	n, err := rand.Int(rand.Reader, big.NewInt(span))
+	if err != nil {
+		return 0, fmt.Errorf("basic: random source: %w", err)
+	}
+	return int(n.Int64()) + min, nil
+}
+
+// randString returns a length-character string drawn uniformly from charset using
+// crypto/rand (no modulo bias).
+func randString(length int, charset string) (string, error) {
+	if length < 1 || length > maxRandomLen {
+		return "", fmt.Errorf("basic: length must be between 1 and %d: %w", maxRandomLen, ErrInvalidArgument)
+	}
+	cs := []rune(charset)
+	if len(cs) == 0 {
+		return "", fmt.Errorf("basic: charset must not be empty: %w", ErrInvalidArgument)
+	}
+	var b strings.Builder
+	b.Grow(length)
+	span := big.NewInt(int64(len(cs)))
+	for i := 0; i < length; i++ {
+		n, err := rand.Int(rand.Reader, span)
+		if err != nil {
+			return "", fmt.Errorf("basic: random source: %w", err)
+		}
+		b.WriteRune(cs[n.Int64()])
+	}
+	return b.String(), nil
+}
diff --git a/plugins/tools/basic/random_test.go b/plugins/tools/basic/random_test.go
new file mode 100644
index 0000000..03b0174
--- /dev/null
+++ b/plugins/tools/basic/random_test.go
@@ -0,0 +1,86 @@
+package basic
+
+import (
+	"strings"
+	"testing"
+)
+
+// TestRandom_Execute_Normal covers integer range and string generation.
+func TestRandom_Execute_Normal(t *testing.T) {
+	t.Parallel()
+	r := random{}
+
+	t.Run("integer in range", func(t *testing.T) {
+		t.Parallel()
+		for i := 0; i < 200; i++ {
+			got, err := run(t, r, map[string]any{"min": 5, "max": 9})
+			assertNoErr(t, err)
+			v := fieldInt(t, mapResult(t, got), "value")
+			if v < 5 || v > 9 {
+				t.Fatalf("value %d out of range [5,9]", v)
+			}
+		}
+	})
+
+	t.Run("string length and charset", func(t *testing.T) {
+		t.Parallel()
+		got, err := run(t, r, map[string]any{"length": 12, "charset": "ab"})
+		assertNoErr(t, err)
+		s := fieldString(t, mapResult(t, got), "value")
+		if len([]rune(s)) != 12 {
+			t.Fatalf("length = %d, want 12", len([]rune(s)))
+		}
+		if strings.Trim(s, "ab") != "" {
+			t.Fatalf("value %q contains characters outside charset", s)
+		}
+	})
+
+	t.Run("single value range", func(t *testing.T) {
+		t.Parallel()
+		got, err := run(t, r, map[string]any{"min": 7, "max": 7})
+		assertNoErr(t, err)
+		if v := fieldInt(t, mapResult(t, got), "value"); v != 7 {
+			t.Fatalf("value = %d, want 7", v)
+		}
+	})
+}
+
+// TestRandom_Execute_Errors covers invalid parameters.
+func TestRandom_Execute_Errors(t *testing.T) {
+	t.Parallel()
+	r := random{}
+	cases := []struct {
+		name string
+		args map[string]any
+	}{
+		{name: "no args", args: map[string]any{}},
+		{name: "inverted range", args: map[string]any{"min": 10, "max": 2}},
+		{name: "min not integer", args: map[string]any{"min": "x", "max": 5}},
+		{name: "zero length", args: map[string]any{"length": 0}},
+		{name: "negative length", args: map[string]any{"length": -1}},
+		{name: "length too big", args: map[string]any{"length": maxRandomLen + 1}},
+		{name: "empty charset", args: map[string]any{"length": 4, "charset": ""}},
+		{name: "non-string charset", args: map[string]any{"length": 4, "charset": 9}},
+	}
+	for _, tc := range cases {
+		tc := tc
+		t.Run(tc.name, func(t *testing.T) {
+			t.Parallel()
+			_, err := run(t, r, tc.args)
+			assertIs(t, err, ErrInvalidArgument)
+		})
+	}
+}
+
+// TestRandom_Schema verifies the JSON Schema contract.
+func TestRandom_Schema(t *testing.T) {
+	t.Parallel()
+	p, err := random{}.Parameters()
+	assertNoErr(t, err)
+	props := p["properties"].(map[string]any)
+	for _, field := range []string{"min", "max", "length"} {
+		if props[field].(map[string]any)["type"] != "integer" {
+			t.Fatalf("%s type should be integer", field)
+		}
+	}
+}
diff --git a/plugins/tools/basic/register.go b/plugins/tools/basic/register.go
new file mode 100644
index 0000000..de7fcca
--- /dev/null
+++ b/plugins/tools/basic/register.go
@@ -0,0 +1,37 @@
+package basic
+
+import (
+	"fmt"
+
+	"git.yeij.top/AskaEth/Cyrene_v2/internal/capability/registry"
+	"git.yeij.top/AskaEth/Cyrene_v2/pkg/plugins/sdk"
+)
+
+// Tools returns the built-in basic tools in registration order: calculator,
+// datetime, random, text, json and markdown. Each element is a stateless
+// sdk.Tool value, safe for concurrent use.
+func Tools() []sdk.Tool {
+	return []sdk.Tool{
+		calculator{},
+		datetime{},
+		random{},
+		text{},
+		jsonTool{},
+		markdown{},
+	}
+}
+
+// Register registers every basic tool with reg. It returns an error wrapping
+// sdk.ErrNilRegistrar when reg is nil, and stops at the first failing tool,
+// wrapping the registry error so errors.Is can still reach it.
+func Register(reg *registry.Registry) error {
+	if reg == nil {
+		return fmt.Errorf("basic: register: %w", sdk.ErrNilRegistrar)
+	}
+	for _, t := range Tools() {
+		if err := reg.RegisterTool(t); err != nil {
+			return fmt.Errorf("basic: register %q: %w", t.Name(), err)
+		}
+	}
+	return nil
+}
diff --git a/plugins/tools/basic/register_test.go b/plugins/tools/basic/register_test.go
new file mode 100644
index 0000000..71f1015
--- /dev/null
+++ b/plugins/tools/basic/register_test.go
@@ -0,0 +1,77 @@
+package basic
+
+import (
+	"testing"
+
+	"git.yeij.top/AskaEth/Cyrene_v2/internal/capability/registry"
+	"git.yeij.top/AskaEth/Cyrene_v2/pkg/plugins/sdk"
+)
+
+// TestTools verifies Tools returns six valid, uniquely-named sdk.Tool values.
+func TestTools(t *testing.T) {
+	t.Parallel()
+	tools := Tools()
+	if len(tools) != 6 {
+		t.Fatalf("Tools() = %d tools, want 6", len(tools))
+	}
+	seen := map[string]bool{}
+	for _, tool := range tools {
+		if tool.Name() == "" || tool.Description() == "" {
+			t.Fatalf("tool %q missing Name or Description", tool.Name())
+		}
+		params, err := tool.Parameters()
+		if err != nil {
+			t.Fatalf("tool %q Parameters() error: %v", tool.Name(), err)
+		}
+		if params == nil {
+			t.Fatalf("tool %q Parameters() returned nil", tool.Name())
+		}
+		if seen[tool.Name()] {
+			t.Fatalf("duplicate tool name %q", tool.Name())
+		}
+		seen[tool.Name()] = true
+	}
+}
+
+// TestRegister_Success verifies every tool is registered as a KindTool capability.
+func TestRegister_Success(t *testing.T) {
+	t.Parallel()
+	reg := registry.New()
+	if err := Register(reg); err != nil {
+		t.Fatalf("Register() error: %v", err)
+	}
+	caps := reg.List(registry.KindTool)
+	if len(caps) != 6 {
+		t.Fatalf("List(KindTool) = %d, want 6", len(caps))
+	}
+	for _, name := range []string{"calculator", "datetime", "random", "text", "json", "markdown"} {
+		cap, err := reg.Get(name)
+		if err != nil {
+			t.Fatalf("Get(%q) error: %v", name, err)
+		}
+		if cap.Kind != registry.KindTool {
+			t.Fatalf("capability %q kind = %v, want tool", name, cap.Kind)
+		}
+		if _, ok := cap.Value.(sdk.Tool); !ok {
+			t.Fatalf("capability %q not an sdk.Tool", name)
+		}
+	}
+}
+
+// TestRegister_NilRegistry verifies nil registry is rejected.
+func TestRegister_NilRegistry(t *testing.T) {
+	t.Parallel()
+	err := Register(nil)
+	assertIs(t, err, sdk.ErrNilRegistrar)
+}
+
+// TestRegister_Duplicate verifies a second registration is rejected.
+func TestRegister_Duplicate(t *testing.T) {
+	t.Parallel()
+	reg := registry.New()
+	if err := Register(reg); err != nil {
+		t.Fatalf("first Register() error: %v", err)
+	}
+	err := Register(reg)
+	assertIs(t, err, registry.ErrDuplicate)
+}
diff --git a/plugins/tools/basic/text.go b/plugins/tools/basic/text.go
new file mode 100644
index 0000000..cb66445
--- /dev/null
+++ b/plugins/tools/basic/text.go
@@ -0,0 +1,161 @@
+package basic
+
+import (
+	"context"
+	"fmt"
+	"strings"
+	"unicode"
+	"unicode/utf8"
+)
+
+// text exposes common string transformations: case folding, trimming, length,
+// splitting, replacement and template placeholder substitution.
+type text struct{}
+
+// Name returns the tool identifier.
+func (text) Name() string { return "text" }
+
+// Description explains what the tool does for the model.
+func (text) Description() string {
+	return "Transform or measure text: upper, lower, title, trim, length, split, replace and template placeholder substitution."
+}
+
+// Parameters returns the tool input JSON Schema. The operation-specific inputs
+// (delimiter, old/new for replace, values for template) are validated by Execute
+// because they only apply to their operation.
+func (text) Parameters() (map[string]any, error) {
+	return map[string]any{
+		"type": "object",
+		"properties": map[string]any{
+			"operation": map[string]any{
+				"type":        "string",
+				"enum":        []any{"upper", "lower", "title", "trim", "length", "split", "replace", "template"},
+				"description": "the transformation to apply",
+			},
+			"input":     map[string]any{"type": "string", "description": "the input string"},
+			"delimiter": map[string]any{"type": "string", "description": "split delimiter"},
+			"old":       map[string]any{"type": "string", "description": "substring to replace"},
+			"new":       map[string]any{"type": "string", "description": "replacement for old"},
+			"values":    map[string]any{"type": "object", "description": "template placeholder substitutions"},
+		},
+		"required": []any{"operation", "input"},
+	}, nil
+}
+
+// Execute applies the requested operation.
+func (text) Execute(ctx context.Context, args any) (any, error) {
+	if err := ctx.Err(); err != nil {
+		return nil, err
+	}
+	m, err := asMap(args)
+	if err != nil {
+		return nil, err
+	}
+	op, err := reqString(m, "operation")
+	if err != nil {
+		return nil, err
+	}
+	in, err := reqString(m, "input")
+	if err != nil {
+		return nil, err
+	}
+	switch strings.ToLower(strings.TrimSpace(op)) {
+	case "upper":
+		return map[string]any{"result": strings.ToUpper(in)}, nil
+	case "lower":
+		return map[string]any{"result": strings.ToLower(in)}, nil
+	case "title":
+		return map[string]any{"result": title(in)}, nil
+	case "trim":
+		return map[string]any{"result": strings.TrimSpace(in)}, nil
+	case "length":
+		return map[string]any{"result": utf8.RuneCountInString(in)}, nil
+	case "split":
+		delim, err := reqString(m, "delimiter")
+		if err != nil {
+			return nil, err
+		}
+		parts := strings.Split(in, delim)
+		out := make([]any, len(parts))
+		for i, p := range parts {
+			out[i] = p
+		}
+		return map[string]any{"result": out}, nil
+	case "replace":
+		old, err := reqString(m, "old")
+		if err != nil {
+			return nil, err
+		}
+		repl, err := reqString(m, "new")
+		if err != nil {
+			return nil, err
+		}
+		return map[string]any{"result": strings.ReplaceAll(in, old, repl)}, nil
+	case "template":
+		vals, err := reqStringMap(m, "values")
+		if err != nil {
+			return nil, err
+		}
+		out, err := renderTemplate(in, vals)
+		if err != nil {
+			return nil, err
+		}
+		return map[string]any{"result": out}, nil
+	default:
+		return nil, fmt.Errorf("basic: unknown text operation %q: %w", op, ErrUnsupported)
+	}
+}
+
+// title returns the title case of s: the first letter of each whitespace-delimited
+// word upper-cased and the rest lower-cased, preserving the original layout.
+func title(s string) string {
+	var b strings.Builder
+	b.Grow(len(s))
+	atStart := true
+	for _, r := range s {
+		if unicode.IsLetter(r) {
+			if atStart {
+				b.WriteRune(unicode.ToUpper(r))
+			} else {
+				b.WriteRune(unicode.ToLower(r))
+			}
+			atStart = false
+			continue
+		}
+		b.WriteRune(r)
+		if unicode.IsSpace(r) {
+			atStart = true
+		}
+	}
+	return b.String()
+}
+
+// renderTemplate substitutes {key} placeholders in tpl using vals, leaving any
+// placeholder whose key is absent unchanged. It scans once so a value that itself
+// contains a placeholder is not re-substituted.
+func renderTemplate(tpl string, vals map[string]string) (string, error) {
+	var b strings.Builder
+	b.Grow(len(tpl))
+	for i := 0; i < len(tpl); {
+		open := strings.IndexByte(tpl[i:], '{')
+		if open < 0 {
+			b.WriteString(tpl[i:])
+			break
+		}
+		b.WriteString(tpl[i : i+open])
+		close := strings.IndexByte(tpl[i+open:], '}')
+		if close < 0 {
+			return "", fmt.Errorf("basic: unbalanced placeholder in template: %w", ErrInvalidArgument)
+		}
+		start := i + open + 1
+		end := i + open + close
+		key := tpl[start:end]
+		if val, ok := vals[key]; ok {
+			b.WriteString(val)
+		} else {
+			b.WriteString(tpl[i+open : end+1])
+		}
+		i = end + 1
+	}
+	return b.String(), nil
+}
diff --git a/plugins/tools/basic/text_test.go b/plugins/tools/basic/text_test.go
new file mode 100644
index 0000000..6d63eb7
--- /dev/null
+++ b/plugins/tools/basic/text_test.go
@@ -0,0 +1,107 @@
+package basic
+
+import (
+	"reflect"
+	"testing"
+)
+
+// TestText_Execute_Normal covers each supported operation.
+func TestText_Execute_Normal(t *testing.T) {
+	t.Parallel()
+	tool := text{}
+	cases := []struct {
+		name  string
+		args  map[string]any
+		check func(t *testing.T, m map[string]any)
+	}{
+		{name: "upper", args: map[string]any{"operation": "upper", "input": "abc"}, check: func(t *testing.T, m map[string]any) {
+			if got := fieldString(t, m, "result"); got != "ABC" {
+				t.Fatalf("result = %q, want ABC", got)
+			}
+		}},
+		{name: "lower", args: map[string]any{"operation": "lower", "input": "ABC"}, check: func(t *testing.T, m map[string]any) {
+			if got := fieldString(t, m, "result"); got != "abc" {
+				t.Fatalf("result = %q, want abc", got)
+			}
+		}},
+		{name: "trim", args: map[string]any{"operation": "trim", "input": "  x  "}, check: func(t *testing.T, m map[string]any) {
+			if got := fieldString(t, m, "result"); got != "x" {
+				t.Fatalf("result = %q, want x", got)
+			}
+		}},
+		{name: "title", args: map[string]any{"operation": "title", "input": "hello world"}, check: func(t *testing.T, m map[string]any) {
+			if got := fieldString(t, m, "result"); got != "Hello World" {
+				t.Fatalf("result = %q, want Hello World", got)
+			}
+		}},
+		{name: "length", args: map[string]any{"operation": "length", "input": "héllo"}, check: func(t *testing.T, m map[string]any) {
+			if got := fieldInt(t, m, "result"); got != 5 {
+				t.Fatalf("result = %d, want 5 runes", got)
+			}
+		}},
+		{name: "split", args: map[string]any{"operation": "split", "input": "a,b,c", "delimiter": ","}, check: func(t *testing.T, m map[string]any) {
+			got := m["result"].([]any)
+			if !reflect.DeepEqual(got, []any{"a", "b", "c"}) {
+				t.Fatalf("result = %v, want [a b c]", got)
+			}
+		}},
+		{name: "replace", args: map[string]any{"operation": "replace", "input": "a-b-a", "old": "a", "new": "x"}, check: func(t *testing.T, m map[string]any) {
+			if got := fieldString(t, m, "result"); got != "x-b-x" {
+				t.Fatalf("result = %q, want x-b-x", got)
+			}
+		}},
+		{name: "template", args: map[string]any{"operation": "template", "input": "Hello {name}!", "values": map[string]any{"name": "Ada"}}, check: func(t *testing.T, m map[string]any) {
+			if got := fieldString(t, m, "result"); got != "Hello Ada!" {
+				t.Fatalf("result = %q, want Hello Ada!", got)
+			}
+		}},
+	}
+	for _, tc := range cases {
+		tc := tc
+		t.Run(tc.name, func(t *testing.T) {
+			t.Parallel()
+			got, err := run(t, tool, tc.args)
+			assertNoErr(t, err)
+			tc.check(t, mapResult(t, got))
+		})
+	}
+}
+
+// TestText_Execute_Errors covers unknown operations and missing per-op inputs.
+func TestText_Execute_Errors(t *testing.T) {
+	t.Parallel()
+	tool := text{}
+	cases := []struct {
+		name   string
+		args   map[string]any
+		target error
+	}{
+		{name: "unknown operation", args: map[string]any{"operation": "frobnicate", "input": "x"}, target: ErrUnsupported},
+		{name: "missing input", args: map[string]any{"operation": "upper"}, target: ErrInvalidArgument},
+		{name: "non-string input", args: map[string]any{"operation": "upper", "input": 9}, target: ErrInvalidArgument},
+		{name: "split without delimiter", args: map[string]any{"operation": "split", "input": "a"}, target: ErrInvalidArgument},
+		{name: "replace without old", args: map[string]any{"operation": "replace", "input": "a", "new": "b"}, target: ErrInvalidArgument},
+		{name: "replace without new", args: map[string]any{"operation": "replace", "input": "a", "old": "a"}, target: ErrInvalidArgument},
+		{name: "template without values", args: map[string]any{"operation": "template", "input": "{x}"}, target: ErrInvalidArgument},
+		{name: "unbalanced placeholder", args: map[string]any{"operation": "template", "input": "{x", "values": map[string]any{}}, target: ErrInvalidArgument},
+	}
+	for _, tc := range cases {
+		tc := tc
+		t.Run(tc.name, func(t *testing.T) {
+			t.Parallel()
+			_, err := run(t, tool, tc.args)
+			assertIs(t, err, tc.target)
+		})
+	}
+}
+
+// TestText_Schema verifies required fields.
+func TestText_Schema(t *testing.T) {
+	t.Parallel()
+	p, err := text{}.Parameters()
+	assertNoErr(t, err)
+	reqs := p["required"].([]any)
+	if len(reqs) != 2 || reqs[0] != "operation" || reqs[1] != "input" {
+		t.Fatalf("required = %v, want [operation input]", reqs)
+	}
+}
diff --git a/plugins/tools/basic/util.go b/plugins/tools/basic/util.go
new file mode 100644
index 0000000..4515396
--- /dev/null
+++ b/plugins/tools/basic/util.go
@@ -0,0 +1,192 @@
+package basic
+
+import (
+	"encoding/json"
+	"errors"
+	"fmt"
+	"math"
+)
+
+// Sentinel errors returned by the basic tools. Use errors.Is to test for them.
+var (
+	// ErrInvalidArgument is returned when a tool is given a missing required
+	// argument or an argument of the wrong type.
+	ErrInvalidArgument = errors.New("basic: invalid argument")
+	// ErrUnsupported is returned when an operation is not in the tool allowlist.
+	ErrUnsupported = errors.New("basic: unsupported operation")
+	// ErrEvaluation is returned when a calculator expression cannot be parsed or
+	// evaluated.
+	ErrEvaluation = errors.New("basic: evaluation failed")
+)
+
+// asMap asserts args is a nil value or a map[string]any — the shape the
+// dispatcher passes when a tool's Parameters declares an object schema.
+func asMap(args any) (map[string]any, error) {
+	if args == nil {
+		return map[string]any{}, nil
+	}
+	m, ok := args.(map[string]any)
+	if !ok {
+		return nil, fmt.Errorf("basic: arguments must be an object: %w", ErrInvalidArgument)
+	}
+	return m, nil
+}
+
+// hasKey reports whether m contains key.
+func hasKey(m map[string]any, key string) bool {
+	_, ok := m[key]
+	return ok
+}
+
+// reqString returns the string argument under key, erroring when it is absent or
+// not a string.
+func reqString(m map[string]any, key string) (string, error) {
+	v, ok := m[key]
+	if !ok {
+		return "", fmt.Errorf("basic: missing required argument %q: %w", key, ErrInvalidArgument)
+	}
+	s, ok := v.(string)
+	if !ok {
+		return "", fmt.Errorf("basic: argument %q must be a string: %w", key, ErrInvalidArgument)
+	}
+	return s, nil
+}
+
+// optString returns the string argument under key, or def when absent. A present
+// non-string is an error.
+func optString(m map[string]any, key, def string) (string, error) {
+	if !hasKey(m, key) {
+		return def, nil
+	}
+	return reqString(m, key)
+}
+
+// reqInt returns the integer argument under key, erroring when it is absent, not
+// a number, or not integral.
+func reqInt(m map[string]any, key string) (int, error) {
+	v, ok := m[key]
+	if !ok {
+		return 0, fmt.Errorf("basic: missing required argument %q: %w", key, ErrInvalidArgument)
+	}
+	n, err := asInt(v)
+	if err != nil {
+		return 0, fmt.Errorf("basic: argument %q: %w", key, err)
+	}
+	return n, nil
+}
+
+// optInt returns the integer argument under key, or def when absent.
+func optInt(m map[string]any, key string, def int) (int, error) {
+	if !hasKey(m, key) {
+		return def, nil
+	}
+	return reqInt(m, key)
+}
+
+// optBool returns the boolean argument under key, or def when absent.
+func optBool(m map[string]any, key string, def bool) (bool, error) {
+	if !hasKey(m, key) {
+		return def, nil
+	}
+	v, ok := m[key].(bool)
+	if !ok {
+		return false, fmt.Errorf("basic: argument %q must be a boolean: %w", key, ErrInvalidArgument)
+	}
+	return v, nil
+}
+
+// reqStringMap returns the object argument under key coerced to map[string]string,
+// erroring when a value is not a string.
+func reqStringMap(m map[string]any, key string) (map[string]string, error) {
+	v, ok := m[key]
+	if !ok {
+		return nil, fmt.Errorf("basic: missing required argument %q: %w", key, ErrInvalidArgument)
+	}
+	raw, ok := v.(map[string]any)
+	if !ok {
+		return nil, fmt.Errorf("basic: argument %q must be an object: %w", key, ErrInvalidArgument)
+	}
+	out := make(map[string]string, len(raw))
+	for k, val := range raw {
+		s, ok := val.(string)
+		if !ok {
+			return nil, fmt.Errorf("basic: value %q must be a string: %w", k, ErrInvalidArgument)
+		}
+		out[k] = s
+	}
+	return out, nil
+}
+
+// strSliceArg returns the array-of-strings argument under key.
+func strSliceArg(m map[string]any, key string) ([]string, error) {
+	v, ok := m[key]
+	if !ok {
+		return nil, fmt.Errorf("basic: missing required argument %q: %w", key, ErrInvalidArgument)
+	}
+	raw, ok := v.([]any)
+	if !ok {
+		return nil, fmt.Errorf("basic: argument %q must be an array: %w", key, ErrInvalidArgument)
+	}
+	out := make([]string, len(raw))
+	for i, item := range raw {
+		s, ok := item.(string)
+		if !ok {
+			return nil, fmt.Errorf("basic: argument %q item %d must be a string: %w", key, i, ErrInvalidArgument)
+		}
+		out[i] = s
+	}
+	return out, nil
+}
+
+// strSliceSliceArg returns the array-of-arrays-of-strings argument under key.
+func strSliceSliceArg(m map[string]any, key string) ([][]string, error) {
+	v, ok := m[key]
+	if !ok {
+		return nil, fmt.Errorf("basic: missing required argument %q: %w", key, ErrInvalidArgument)
+	}
+	raw, ok := v.([]any)
+	if !ok {
+		return nil, fmt.Errorf("basic: argument %q must be an array: %w", key, ErrInvalidArgument)
+	}
+	out := make([][]string, len(raw))
+	for i, item := range raw {
+		row, ok := item.([]any)
+		if !ok {
+			return nil, fmt.Errorf("basic: argument %q item %d must be an array: %w", key, i, ErrInvalidArgument)
+		}
+		cells := make([]string, len(row))
+		for j, c := range row {
+			s, ok := c.(string)
+			if !ok {
+				return nil, fmt.Errorf("basic: argument %q item %d cell %d must be a string: %w", key, i, j, ErrInvalidArgument)
+			}
+			cells[j] = s
+		}
+		out[i] = cells
+	}
+	return out, nil
+}
+
+// asInt converts a numeric-typed value (the types the dispatcher decodes JSON
+// into) to an int, rejecting non-integral floats.
+func asInt(v any) (int, error) {
+	switch x := v.(type) {
+	case int:
+		return x, nil
+	case int64:
+		return int(x), nil
+	case float64:
+		if x != math.Trunc(x) || x < math.MinInt64 || x > math.MaxInt64 {
+			return 0, fmt.Errorf("must be an integer: %w", ErrInvalidArgument)
+		}
+		return int(x), nil
+	case json.Number:
+		n, err := x.Int64()
+		if err != nil {
+			return 0, fmt.Errorf("must be an integer: %w", ErrInvalidArgument)
+		}
+		return int(n), nil
+	default:
+		return 0, fmt.Errorf("must be a number: %w", ErrInvalidArgument)
+	}
+}