feat(policy): enforce truthfulness and autonomy guardrails

Add runtime truthfulness modes and autonomy-level tool gating with audit metadata for overrides/denials.

Wire policy through prompt assembly, tool execution context, and daemon/gateway agent paths; update tests and planning state for Phase 3 PR #2 completion.
This commit is contained in:
William Valentin
2026-02-12 16:06:45 -08:00
parent 125af4e832
commit 90ce622080
18 changed files with 1172 additions and 104 deletions
+37
View File
@@ -35,6 +35,13 @@ const bigOutputTool: Tool = {
execute: async () => ({ success: true, output: 'x'.repeat(100_000) }),
};
const fileWriteLikeTool: Tool = {
name: 'file.write',
description: 'Test file write tool',
inputSchema: { type: 'object', properties: {} },
execute: async () => ({ success: true, output: 'ok' }),
};
describe('ToolExecutor', () => {
it('executes a tool and returns result', async () => {
const registry = new ToolRegistry();
@@ -123,4 +130,34 @@ describe('ToolExecutor', () => {
expect(result.success).toBe(false);
expect(result.error).toContain('denied');
});
it('conservative autonomy requires confirm for dangerous tools', async () => {
const registry = new ToolRegistry();
registry.register(fileWriteLikeTool);
const hooks = new HookEngine({ confirm: [], log: [], silent: [] });
const executor = new ToolExecutor(registry, hooks);
const resultPromise = executor.execute('file.write', {}, { autonomyLevel: 'conservative' });
const pending = hooks.getPendingConfirmations();
expect(pending).toHaveLength(1);
hooks.resolveConfirmation(pending[0].id, { approved: true });
const result = await resultPromise;
expect(result.success).toBe(true);
});
it('autonomous mode defers to explicit confirm hooks', async () => {
const registry = new ToolRegistry();
registry.register(echoTool);
const hooks = new HookEngine({ confirm: ['test.*'], log: [], silent: [] });
const executor = new ToolExecutor(registry, hooks);
const resultPromise = executor.execute('test.echo', { text: 'hi' }, { autonomyLevel: 'autonomous' });
const pending = hooks.getPendingConfirmations();
expect(pending).toHaveLength(1);
hooks.resolveConfirmation(pending[0].id, { approved: true });
const result = await resultPromise;
expect(result.success).toBe(true);
});
});