mirror of
https://github.com/garrytan/gstack.git
synced 2026-11-05 00:03:15 +01:00
3b7254a54d
Outside-review consent (codex_reviews) is enforced in gstack-codex-probe run-with-timeout, gstack-claude-code, gstack-outside-voice run and the autoplan runner (not_run / consent_missing, exit 3; #3102). Codex reasoning effort resolves GSTACK_CODEX_EFFORT, then codex_effort, then each gate's default inside run-with-timeout --effort (D1-A; #3098, closes #2934). README and telemetry privacy disclosures corrected (#1081, #1080, #1150). Co-authored-by: Lockyer <135391289+Lockyer228@users.noreply.github.com> Co-authored-by: Lubos Buracinsky <159481718+lubosxyz@users.noreply.github.com>
71 lines
3.5 KiB
TypeScript
71 lines
3.5 KiB
TypeScript
/**
|
|
* #2847: a nested gstack Codex call is a one-shot review, but Codex injects every
|
|
* installed skill (gstack's own `review` included) into its context, and the
|
|
* reviewer re-ran whole skill workflows inside its timeout. A read-only sandbox
|
|
* does not stop that. Verified against codex-cli 0.152.0 and 0.160.0 with a local
|
|
* Responses API stub: without `skills.include_instructions=false` the request
|
|
* carried the full gstack catalog; with it, none (exec and native review alike).
|
|
* Every gstack-owned Codex call therefore carries the isolation flag, and the
|
|
* prompt boundary names the Codex skill roots.
|
|
*/
|
|
import { describe, expect, test } from 'bun:test';
|
|
import * as fs from 'fs';
|
|
import * as path from 'path';
|
|
import { CODEX_MODEL_CONFIG_FLAG, CODEX_REVIEW_MODEL_CONFIG_FLAG, CODEX_SKILLS_ISOLATION_FLAG } from '../scripts/resolvers/constants';
|
|
import { expectMentions } from './helpers/prompt-structure';
|
|
|
|
const ROOT = path.join(import.meta.dir, '..');
|
|
|
|
function renderedSkillFiles(): string[] {
|
|
const files: string[] = [];
|
|
const pending = [ROOT];
|
|
const excluded = new Set(['node_modules', '.git', '.context', '.claude', 'fixtures']);
|
|
while (pending.length) {
|
|
const dir = pending.pop()!;
|
|
for (const entry of fs.readdirSync(dir, { withFileTypes: true })) {
|
|
if (excluded.has(entry.name)) continue;
|
|
const full = path.join(dir, entry.name);
|
|
if (entry.isDirectory()) pending.push(full);
|
|
else if (entry.name.endsWith('.md') && !entry.name.endsWith('.tmpl')) files.push(full);
|
|
}
|
|
}
|
|
return files;
|
|
}
|
|
|
|
describe('Codex skill isolation (#2847)', () => {
|
|
test('the model flags carry the isolation flag, so every model-flag site is isolated', () => {
|
|
expect(CODEX_SKILLS_ISOLATION_FLAG).toBe('-c skills.include_instructions=false');
|
|
expect(CODEX_MODEL_CONFIG_FLAG).toContain(CODEX_SKILLS_ISOLATION_FLAG);
|
|
expect(CODEX_REVIEW_MODEL_CONFIG_FLAG).toContain(CODEX_SKILLS_ISOLATION_FLAG);
|
|
});
|
|
|
|
test('every wrapped Codex exec/review call in generated skills disables the skill catalog', () => {
|
|
const calls: string[] = [];
|
|
for (const file of renderedSkillFiles()) {
|
|
for (const line of fs.readFileSync(file, 'utf8').split('\n')) {
|
|
if (/run-with-timeout\s+(?:--user-invoked\s+)?(?:--effort\s+\w+\s+)?\d+\s+codex\s+(exec|review)\b/.test(line)) calls.push(`${path.relative(ROOT, file)}: ${line}`);
|
|
}
|
|
}
|
|
expect(calls.length).toBeGreaterThan(10);
|
|
expect(calls.filter(call => !call.includes("skills.include_instructions=false"))).toEqual([]);
|
|
});
|
|
|
|
test('the probe round trip does not pay for the skill catalog either', () => {
|
|
const probe = fs.readFileSync(path.join(ROOT, 'bin', 'gstack-codex-probe'), 'utf8');
|
|
const roundTrip = probe.split('\n').find(line => line.includes('"reply OK"'));
|
|
expect(roundTrip).toContain("skills.include_instructions=false");
|
|
});
|
|
|
|
test('outside-voice and /codex prompt boundaries name the Codex skill roots and forbid skill use', () => {
|
|
const sources = ['scripts/resolvers/outside-voice-steps.ts', ...['review-mode', 'challenge-mode', 'consult-mode'].map(m => `codex/sections/${m}.md.tmpl`)];
|
|
for (const rel of sources) {
|
|
const text = fs.readFileSync(path.join(ROOT, rel), 'utf8');
|
|
const boundaries = text.split('\n').filter(line => /do not read or execute any files under/i.test(line));
|
|
expect(boundaries.length, rel).toBeGreaterThan(0);
|
|
for (const line of boundaries) {
|
|
expectMentions(line, [['do not', 'installed', 'invoke']], 'line');
|
|
}
|
|
}
|
|
});
|
|
});
|