Skill Claude CodeCodex
Bootstrap or upgrade the project-engineering-workflow framework for a repository. Use when the user explicitly wants to add AGENTS plus local skills plus spec artifacts to a repo, audit or upgrade an existing project-engineering-workflow installation, or set up this specific engineering workflow starter. Do not use…
Skill Claude CodeCodex
Govern feature branches, release branches, version tags, and merge-to-main flow. Use when starting tracked development, preparing a release, creating tags, or deciding whether a change is ready to merge into main.
Skill Claude CodeCodex
Implement the confirmed plan with minimal safe edits. Use after requirement, onboarding, and scope are clear so Codex can change code while preserving repository conventions and reuse.
Skill Claude CodeCodex
Review project changes for bugs, regressions, unsafe assumptions, and missing tests. Use before merge, handoff, or final delivery.
Skill Claude CodeCodex
Build a read-only codebase map before editing unfamiliar areas. Use when Codex is entering a new module, inherited project, or complex repository and must explain request flow, ownership boundaries, source of truth, reuse points, and likely risks.
Skill Claude CodeCodex
Upgrade the workflow from task learnings. Use when Codex should convert repeated session feedback, rule drift, process friction, recurring failures, or confirmed best practices into proposed changes for AGENTS, skills, templates, constitution, or memory policy.
Skill Claude CodeCodex
Orchestrate long-running or context-heavy work using a Get Stuff Done style workflow. Use when a task spans many files, multiple sessions, multiple milestones, unclear sequencing, or needs resumable execution state under specs/ /.
Skill Claude CodeCodex
Route role-based product, design, engineering, QA, release, and reflection decisions. Use when a task needs explicit role separation, product judgment, design review, engineering tradeoff review, QA gates, ship readiness, or post-delivery reflection.
Skill Claude CodeCodex
Govern project memory safely. Use when a request asks Codex to remember, forget, update, retrieve, summarize, personalize, share team knowledge, preserve session learnings, resolve memory conflicts, or decide whether information should become durable memory.
Skill Claude CodeCodex
Analyze project requests before implementation. Use when a user describes a feature, bug, refactor, integration, or workflow update and Codex must first summarize the request in Chinese, list constraints, define acceptance criteria, identify impacted modules, and highlight risky assumptions.
Skill Claude CodeCodex
Lock the smallest safe change scope before implementation. Use when a confirmed request may trigger cross-file or cross-layer edits and Codex must separate direct impact from ripple impact, define out-of-scope items, and prevent accidental overreach.
Skill Claude CodeCodex
Implement changes using the repository's actual stack conventions. Use when editing the main application under APPPATH and Codex needs stack-specific guidance for STACKNAME.
Skill Claude CodeCodex
Route high-leverage AI capabilities safely inside this project. Use when a request may benefit from browser automation, image/document/spreadsheet/presentation tools, multi-agent research, GSD long-task orchestration, gstack role review, recurring checks, external APIs, generated assets, or other enhanced execution…
Skill Claude CodeCodex
Convert a confirmed request into an implementation plan. Use for cross-layer features, risky bug fixes, integrations, refactors, or any change that benefits from a written technical solution before coding.
Skill Claude CodeCodex
Validate project changes before delivery. Use after implementation to run low-cost automated checks, summarize manual verification steps, and provide a Chinese test report.
At most 3 mods per repository are shown here, and a mod shipped inside a plugin is left to that plugin's page — the rest are on their repository pages: