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/lib/events.py

View on GitHub
"""Read executor envelopes, never count an event-shaped string inside assistant prose."""
import json
from pathlib import Path
import sys


def inspect(path):
    events = [json.loads(line) for line in Path(path).read_text().splitlines() if line.strip()]
    commands = 0
    completed = False
    failure = None
    for e in events:
        # An error frame is not by itself a failed round. Measured: a transport dropped mid-run,
        # logged four reconnect errors, recovered, and closed a 58KB inventory with `turn.completed`
        # as the last event and exit 0 -- and this refused it as "executor reported failure". The
        # file distinguishes them without loosening anything: an error that is followed by a
        # completion was survived, and an error that ends the stream was not. So the failure is
        # remembered rather than raised, and only stands if nothing completes after it.
        if e.get('type') in {'turn.failed', 'error'}:
            failure = 'executor reported failure'
            completed = False
        if e.get('type') == 'turn.started':
            completed = False
        if e.get('type') == 'turn.completed':
            completed, failure = True, None
        if e.get('type') == 'result':
            if e.get('is_error') or e.get('subtype', 'success') != 'success':
                raise ValueError('executor result was not successful')
            completed, failure = True, None
        item = e.get('item') or {}
        if e.get('type') == 'item.completed' and item.get('type') == 'command_execution':
            commands += 1
        if e.get('type') == 'assistant':
            commands += sum(c.get('type') == 'tool_use' for c in
                            (e.get('message') or {}).get('content', []))
    # Stands only if nothing completed after it: an error that ends the stream failed the round,
    # an error the round outlived did not.
    if failure:
        raise ValueError(failure)
    return completed, commands


if __name__ == '__main__':
    try:
        completed, count = inspect(sys.argv[2])
        if sys.argv[1] == 'completed' and not completed:
            raise ValueError('no successful completion event')
        if sys.argv[1] == 'commands' and not count:
            raise ValueError('reviewer executed 0 tools')
        print(count)
    except (ValueError, OSError, TypeError, AttributeError) as e:
        sys.exit(f'GUARD FAIL [events-{sys.argv[1]}] {e}')
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