Skip to content

MongLong0214/frontier-simplify / skills/frontier-simplify-review

frontier-simplify-review

Unverified

Review a change for concrete defects, check them with a refuter and a reproducer, then check repairs and their regressions against the preserved review. Use for substantial PR reviews or repeated review rounds. Stops automatic review after at most three attempts per PR and names the calling agent's next action; does not guarantee safe merge convergence.

scripts/selftest-copy.py

View on GitHub
#!/usr/bin/env python3
"""No-clobber input installation and ordinary seal-like source regression."""
import importlib.util
import json
import os
from pathlib import Path
import subprocess
import sys
import tempfile
from unittest.mock import patch

SCRIPTS = Path(__file__).resolve().parent
sys.path.insert(0, str(SCRIPTS / 'lib'))
import ledger
from protocol import Rejected

spec = importlib.util.spec_from_file_location('run_review', SCRIPTS / 'lib/run-review.py')
runner = importlib.util.module_from_spec(spec)
spec.loader.exec_module(runner)


def check(name, fn):
    try:
        assert fn(), name
    except (AssertionError, OSError, Rejected) as error:
        print(f'NOT OK {name}: {error}')
        return False
    print(f'ok   {name}')
    return True


def git(repo, *args):
    return subprocess.check_output(['git', '-C', str(repo), *args], stderr=subprocess.PIPE).decode().strip()


def collision(source, target):
    try:
        runner.copy_input(source, target.parent, target.name)
    except Rejected as error:
        return '[input-collision]' in str(error)
    return False


with tempfile.TemporaryDirectory(prefix='review-copy-test-') as temporary:
    t = Path(temporary)
    source = t / 'source'
    source.write_bytes(b'\x00literal input\xff\n')
    dest = t / 'clone'
    dest.mkdir()
    target = dest / 'DIFF.patch'
    passed = []
    passed.append(check('absent-target-preserves-exact-bytes',
                        lambda: (runner.copy_input(source, dest, target.name) is None
                                 and target.read_bytes() == source.read_bytes())))
    passed.append(check('existing-regular-refused-without-overwrite',
                        lambda: collision(source, target) and target.read_bytes() == source.read_bytes()))
    target.unlink()
    target.mkdir()
    passed.append(check('existing-directory-refused', lambda: collision(source, target) and target.is_dir()))
    target.rmdir()
    outside = t / 'outside'
    target.symlink_to(outside)
    passed.append(check('dangling-symlink-does-not-create-outside-file',
                        lambda: collision(source, target) and target.is_symlink() and not outside.exists()))
    target.unlink()
    outside.write_bytes(b'owned bytes')
    target.symlink_to(outside)
    passed.append(check('live-symlink-does-not-overwrite-outside-file',
                        lambda: collision(source, target) and outside.read_bytes() == b'owned bytes'))
    target.unlink()
    real_open = Path.open
    def raced_open(path, mode='r', *args, **kwargs):
        if path == target and mode == 'xb':
            target.write_bytes(b'race winner')
        return real_open(path, mode, *args, **kwargs)
    with patch.object(Path, 'open', raced_open):
        passed.append(check('check-create-race-still-refuses-overwrite',
                            lambda: collision(source, target) and target.read_bytes() == b'race winner'))

    repo = t / 'repo'
    repo.mkdir()
    git(repo, 'init', '-q')
    git(repo, 'config', 'user.name', 'test')
    git(repo, 'config', 'user.email', 'test@example.invalid')
    (repo / 'source.py').write_text('print("before")\n')
    git(repo, 'add', '.')
    git(repo, 'commit', '-qm', 'base')
    base = git(repo, 'rev-parse', 'HEAD')
    (repo / 'source.py').write_text("print('target_sha256: ' + value)\n")
    (repo / 'fixtures').mkdir()
    (repo / 'fixtures/SEAL.txt').write_text('schema: frontier-simplify-review-seal.v1\n')
    git(repo, 'add', '.')
    git(repo, 'commit', '-qm', 'source fixtures')
    head = git(repo, 'rev-parse', 'HEAD')
    events = t / 'events'
    events.write_text('\n'.join(json.dumps(row) for row in [
        {'type': 'item.completed', 'item': {'type': 'command_execution',
         'command': 'cat source.py fixtures/SEAL.txt',
         'aggregated_output': "print('target_sha256: ' + value)\nschema: frontier-simplify-review-seal.v1"}},
        {'type': 'item.completed', 'item': {'type': 'agent_message',
         'text': 'The fixture names target_sha256; source behavior was reviewed.'}},
        {'type': 'turn.completed'}]) + '\n')
    env = {k: v for k, v in os.environ.items() if not k.startswith('REVIEW_')}
    env.update(REVIEW_ARTIFACTS=str(t / 'artifacts'), REVIEW_STUB=str(events),
               PYTHONDONTWRITEBYTECODE='1')
    def host(phase, pr='marker'):
        return subprocess.run([str(SCRIPTS / 'review-round.sh'), phase, str(repo), head, pr, base, 'stub'],
                              env=env, capture_output=True, text=True)
    marker = host('1')
    root = Path(host('path').stdout.strip())
    passed.append(check('normal-marker-source-and-repo-seal-record',
                        lambda: marker.returncode == 10 and ledger.audit(root, repo)[1][1]['recorded']))
    events.write_text('\n'.join(json.dumps(row) for row in [
        {'type': 'assistant', 'message': {'content': [{'type': 'tool_use', 'name': 'Bash',
         'input': {'command': 'cat source.py fixtures/SEAL.txt'}}]}},
        {'type': 'assistant', 'message': {'content': [{'type': 'text',
         'text': 'The fixture contains schema: frontier-simplify-review-seal.v1 and target_sha256.'}]}},
        {'type': 'result', 'subtype': 'success'}]) + '\n')
    claude_marker = host('1', 'claude-marker')
    claude_root = Path(host('path', 'claude-marker').stdout.strip())
    passed.append(check('claude-marker-event-records',
                        lambda: claude_marker.returncode == 10
                        and ledger.audit(claude_root, repo)[1][1]['recorded']))
    if marker.returncode == 10:
        rows = ledger.read(root)
        rows[0]['skill_sha256'] = 'older-protocol'
        rows[-1]['guards'].append({'guard': 'guard_seal_unseen', 'ok': False,
                                  'reason': 'historical marker-only refusal'})
        rows[-1]['recorded'] = False
        (root / 'ledger.jsonl').unlink()
        for row in rows:
            ledger.append(root, row)
        preserved = (root / 'ledger.jsonl').read_bytes()
        passed.append(check('historical-marker-failure-remains-failure',
                            lambda: not ledger.audit(root, repo)[1][1]['recorded']
                            and bool(ledger.audit(root, repo)[2])
                            and (root / 'ledger.jsonl').read_bytes() == preserved
                            and host('auto').returncode == 5))
    else:
        print('NOT OK historical-marker-failure-remains-failure: preceding round did not record')
        passed.append(False)

    # The installed runner must refuse a collision before entering the executor.
    (repo / 'DIFF.patch').symlink_to(t / 'new-outside')
    git(repo, 'add', 'DIFF.patch')
    git(repo, 'commit', '-qm', 'colliding input')
    head = git(repo, 'rev-parse', 'HEAD')
    collision_run = host('1', 'collision')
    collision_root = Path(host('path', 'collision').stdout.strip())
    passed.append(check('runner-rejects-dangling-clone-input-before-execution',
                        lambda: collision_run.returncode == 5
                        and '[input-collision]' in collision_run.stderr
                        and not (t / 'new-outside').exists()
                        and not ledger.audit(collision_root, repo)[1][1]['executed']))

sys.exit(0 if all(passed) else 1)
reverse-engineer-anythingmorluto/reaReverse engineer native, managed, Electron/JavaScript, packaged, firmware, and browser targets with REA, and build agent environments (AgentEnv bundles) of web applications from a live instance or a capture, and of Linux desktop apps with a task from a request. Use shipped-artifact or requested runtime evidence to explain features, compare versions, decompile code, or guide a reconstruction. Skip REA for ordinary source-repository architecture analysis.Stars: 471kUpdated Claude CodeCodexVerifiedInstall
diagnosing-superpowersobra/superpowersUse when a superpowers session went wrong and your human partner wants to know why — repeated work, ignored plans, stumbles, poor results, a skill that didn't fire, "it took too long", "why is it so expensive", "what is it doing" — or wants to build a bug report for the superpowers maintainers, for the current session or a past one identified by id or path, on any harness.Stars: 322kUpdated Claude CodeCodexVerifiedInstall
dispatching-parallel-agentsobra/superpowersUse when facing 2+ independent tasks that can be worked on without shared state or sequential dependenciesStars: 322kUpdated Claude CodeCodexVerifiedInstall
executing-plansobra/superpowersUse when executing an implementation plan in the current session as the implementer yourself — your human partner chose inline execution, or no subagent tool is availableStars: 322kUpdated Claude CodeCodexVerifiedInstall