verification plugins

by Marat Karimov

Catalyst: disciplined agentic development - SDD pipeline with decision-boundary model tiering, spec hardening, premortem gates, verification, and handoffs.

by Gregory Block

The research debate engine for Special Circumstances: an additive blue side builds, a subtractive red side audits and owns the pass/fail gate, an engine runs the mechanics and the bench rules the final compromise. Best-of-N, the Catechism (a worth-our-time decision), and the full record preserved.

by Gabriel Costa

Weave is a dependency-free engineering layer for Claude Code and Codex. It wraps shell commands to strip noisy successful output while always preserving failures in full, keeps a compact .weave/state.md handoff so context survives across sessions and between agents, and applies one consistent minimalism-and-verification policy so agents reuse existing code, prefer native features over new dependencies, and back every completion claim with real evidence.

by canglei9527

AI 项目维护与初始化工作流：修复 Bug、最小修改、验证与记录。Workflows for bug fixes, focused changes, verification, and project bootstrap.

by akbun

Refactoring skills — change code, docs, and config already in use, gate every change behind automated verification, and report only what needs a human decision

by Genesis Skill Set

Genesis harness skill system for Codex test-first, contract-first, memory-aware autonomous workflows.

by vzert

Make any agent write its own grounded goal-spec, verify it with an independent adversary, and close through a fail-open completion gate. Hardens the built-in /goal.

by Grimaldo Stanzani

Superpowers-derived process disciplines in a calibrated register: fit-ranked tool dispatch, capacity dispatch (task-to-model/effort routing on a calibrated rubric), calibration-first skill authoring, red-green TDD, root-cause-first debugging, design-before-code brainstorming, evidence-before-claims verification, technically rigorous review reception, and midweight planned execution (executable-cold plans + per-task subagent review loop) — deduplicated against the craft-collection toolkit and eval-gated from birth.

by Christopher Boone

Deliberately break code to prove a test can see it: plant the defect an instrument claims to catch, confirm it goes red, and record it so it regresses.

by Christopher Boone

Deliberately break code to prove a test can see it: plant the defect an instrument claims to catch, confirm it goes red, and record it so it regresses.

by Paulo Soares

Runtime-independent, spec-driven development with operational health, audit, maintenance, PWDEV Code migration, isolated autonomous fleets, and guarded external CLI delegation.

by Paulo Soares

Disciplined, approval-gated development: product requirements, executable roadmap, brainstorm-gated plans, subagent-driven execution with a ledger and bounded fix loop, adversarial verification, and isolated cmux fleets.

by Paulo Soares

Portable quality assurance workflows, specialist guidance, evidence handling, and verifiable HTML and PDF reporting for Claude Code, Codex, and Hermes Agent.

by zy

Project-neutral development, ADR, verification, retirement, and risk-routed review workflows for Claude Code.

by Paulo Soares

Portable spec-driven development for Claude Code, Codex, Hermes Agent, and OpenCode with human-readable contracts, durable operational state, traceability, bounded autonomous correction, and isolated fleets.

by Christopher Boone

Write a numbered, resumable checklist for checks run by hand, where every step says what to do, what you should see, and how to tell nothing from broken.

by Christopher Boone

Write a numbered, resumable checklist for checks run by hand, where every step says what to do, what you should see, and how to tell nothing from broken.

by 404.directory

Agent Discovery + Trust: search, verify, and trust tools before calling them — plus webpage understanding and deployment verification.

by 404.directory

Search official AI and cloud documentation, verify public web deployments, and discover trusted MCP tools.

by Long

Lightweight, evidence-first software delivery for Codex. 中文优先的 AI 软件交付工作流。

by David Kooi

Skills from agent-ops-patterns: hand your agent a source (a course, a paper set) and get a disposition record with a quoted file:line or a scoped command with its exit code and output per verdict. First skill: Stanford CS329A "Self-Improving AI Agents".

by Mohammed Talas

Verify-first discipline for Codex. Bundles Codex-native skills (including session handover — a comprehensive handoff so a fresh session can resume without re-deriving or guessing anything) plus an installer for the Codex hook subset that can be enforced today: session/prompt nudges, a codex-availability PATH probe, shell-command guards, Stop guards, task discipline, and speculation checks. Claude Code plugin files are kept separate and unchanged.

by office-skills

Auto Office v3 adaptive orchestration runtime: durable role packets, harness/model/effort routing, adapter trust, quota-aware cost policy, verification, private telemetry, replay, maturity, and isolated self-improvement. Human merge-to-main remains mandatory.

by Michael Kissinger

Spec-anchored workflow for autonomous coding runs on GPT-6 Astra in Codex: brief, run, verify, checkpoint. astra mode has Astra implement each slice; astra-crew routes implementation to gpt-5.6-sol or gpt-5.6-luna subagents by task class. Same spec contract as the Claude Code plugin in this repo.