用 Codex 或 Claude 帮你安装 复制这段 Prompt,粘贴到 Codex、Claude 或其他助手里,让它检查 Skill 页面并帮你完成安装。
直接命令不会经过审查 Prompt;运行前请先检查来源。
npx skills add https://github.com/loomantix/codex-platform --skill feature-dev命令会保持在同一行。复制前请横向滚动并检查完整内容。
想先保存到本地?可下载 SkillsMP 当前能够提供的文件。
PR-first adversarial code review for Codex. Use after implementation or refactorpass on an open draft PR, especially when the user asks to critique, review hard, find bugs, or run the platform review chain. Posts verified findings inline before fixing, then replies and resolves. Lean mode runs the two highest-signal lanes; the deep matrix runs only when the resolved review tier is Deep.
High-fidelity PR-first Codex review chain that opens or reuses a draft PR, records verified findings inline before fixes, and runs critique deep — preceded by refactorpass on this engine's first pass only. Rounds 3+ run in convergence mode. Runs only when the review tier resolved to Deep; hands a Lean changeset back to critique.
Post-push AI review orchestrator for pull requests. Use when the user asks Codex to run AI review on a PR, address Gemini or Copilot findings, dedupe reviewer comments, push fixes, reply in PR threads, or complete the platform review chain. Both modes fire Gemini Flash + Copilot only — no in-skill local review during the loop. Accepts a PR number, optional `deep` (4-iter cap + early-exit on no-fix iters + final `deepcritique`) which asserts the resolved review tier rather than choosing it, and optional `--resume`.
基于 SOC 职业分类
正在显示 SKILL.md
| name | feature-dev |
| description | Guided feature development with codebase understanding and architecture focus |
You are helping a developer implement a new feature. Follow a systematic approach: understand the codebase deeply, identify and ask about all underspecified details, design elegant architectures, then implement.
Goal: Understand what needs to be built
Initial request: $ARGUMENTS
Actions:
Goal: Understand relevant existing code and patterns at both high and low levels
Actions:
Launch 2-3 code-explorer agents in parallel. Each agent should:
Example agent prompts:
Once the agents return, please read all files identified by agents to build deep understanding
Present comprehensive summary of findings and patterns discovered
Goal: Fill in gaps and resolve all ambiguities before designing
CRITICAL: This is one of the most important phases. DO NOT SKIP.
Actions:
If the user says "whatever you think is best", provide your recommendation and get explicit confirmation.
Goal: Design multiple implementation approaches with different trade-offs
Actions:
Goal: Build the feature
DO NOT START WITHOUT USER APPROVAL
Actions:
Goal: Ensure code is simple, DRY, elegant, easy to read, and functionally correct
Actions:
Goal: Document what was accomplished
Actions: