Token导航 LogoToken导航TokenDH.com
研究检索只读github未标认证来源可访问许可证需确认审计通过

forge-reflect-pr锻造反映公关

Agent Skill

forge-reflect-pr 用于查找、检索和筛选相关信息,适合在 Codex、Claude、Cursor、Gemini CLI 中需要根据关键词、任务场景或来源线索快速定位候选结果时使用。可结合来源仓库、安装命令和原始 README 继续核验具体用法。安装前建议确认权限范围、维护状态,以及是否会触发联网、命令执行或文件读写。

总安装

517

周安装

22

GitHub Stars

64

下载量

181
CodexClaudeCursorGemini CLI

安装说明

本站只整理中文说明和来源信息,不托管安装包,也不代用户安装。

GitHub

来源数

2

许可证

unknown

最后核验

2026-05-01

来源状态

来源可访问

安装方式

通过对话安装

复制提示词发给支持本地命令或 Skills 的 AI 助手,先确认命令和权限,再让它执行。

请帮我安装这个 Agent Skill:forge-reflect-pr(锻造反映公关)
来源仓库:https://github.com/mgratzer/forge
仓库路径:skills/forge-reflect-pr
安装命令:
npx skills add https://github.com/mgratzer/forge --skill forge-reflect-pr
安装前请先检查当前环境是否支持对应 CLI,并向我确认将要执行的命令、安装目录、联网范围和文件读写权限;确认后再执行。

命令行安装

复制命令到本机终端执行。该命令会通过 npx skills 从第三方来源获取 Skill;本站只展示命令,不托管安装包,也不自动执行。

skills.shnpx skills
npx skills add https://github.com/mgratzer/forge --skill forge-reflect-pr

简介

用于查找、检索和筛选相关信息,适合在 Codex、Claude、Cursor、Gemini CLI 中快速定位候选结果。

  • 适用于关键词搜索、任务场景匹配或来源线索追踪等研究检索场景。
  • 通过 npx skills add 命令从指定 GitHub 仓库安装并使用。
  • 安装前需确认权限范围、维护状态,注意是否触发联网或文件操作。
  • 建议结合原始 README 和仓库内容进一步验证具体功能和使用方式。

SKILL.md

Reflect on PR

Self-review the current PR branch before requesting peer review.

Input

No primary argument required. Operates on the current branch.

Optional last parameter: -- <additional context>

Interpret $ARGUMENTS as optional execution guidance for the review focus. If no argument is provided, use the default review checklist.

Process

Step 1: Identify Changes

DEFAULT_BRANCH=$(git symbolic-ref refs/remotes/origin/HEAD | sed 's@^refs/remotes/origin/@@')
git diff --name-only $DEFAULT_BRANCH...HEAD

Collect the full diff and the list of changed files for the review step.

Step 2: Review Changes (delegate)

Delegate this step to a sub-agent with fresh context if the runtime supports it. The fresh context eliminates self-review bias — the reviewer has no memory of implementation decisions. If the runtime does not support sub-agents, execute the instructions inline.

Sub-agent instructions:

You are reviewing a PR diff for real problems. Read AGENTS.md to understand project conventions. Classify every finding using the review rubric — only flag P0, P1, and P2 items. For each changed file, check: 1. Duplication — repeated patterns that should be extracted 2. Function size — anything too long to follow at a glance 3. Naming — clear and consistent (only flag if actively misleading, not preference) 4. Layer placement — logic in the right abstraction layer 5. Dead code — unused imports, variables, or functions introduced 6. Pattern consistency — if a pattern was changed, grep for ALL files using it: grep -rn "<changed-pattern>" <search-root>/ 7. Configuration — new env vars documented in sample env or setup docs? Config placed where consumed? No hardcoded credentials? Manual deployment steps captured? 8. Test coverage — corresponding test files exist? Error handling branches covered? Edge cases for new public functions? 9. Documentation — changes require updates to docs/*.md, AGENTS.md, code comments, or README? Grep docs for stale references to anything removed or renamed: grep -rn "<removed-term>" docs/ 10. Cleanup — no temporary debug logging, commented-out code, untracked TODOs, unused imports, or hardcoded values that should be constants Return findings grouped by file, with severity tags (P0/P1/P2).

Inputs provided to sub-agent:

  • Output of git diff $DEFAULT_BRANCH...HEAD
  • List of changed files
  • Contents of review-rubric.md
  • Contents of AGENTS.md (project conventions)
  • Any additional context from the user's invocation

Expected output: Structured findings list with severity tags, grouped by file.

Step 3: Quality Gates

Run the project's lint, format, type check, and test commands. Fix issues and commit fixes.

Step 4: Report

Synthesize the review findings (from Step 2) with quality gate results (from Step 3) into the summary format below.

Step 5: Triage Deferred Items

Present each deferred improvement to the user and ask whether to fix now or defer as a follow-up issue.

For each item, recommend one of:

  • Fix now — small, low-effort changes that fit naturally in this PR (e.g., a missing test case, a stale doc reference, a duplicated line)
  • Defer — larger changes that would expand the PR scope or require separate review (e.g., a cross-cutting refactor, a new feature suggestion)

State your recommendation and let the user decide. Then:

  • Fix now items: apply the fix and commit it
  • Deferred items: create a GitHub issue to track: gh issue create --title "<title>" --body "<context and proposed solution>"

Output Format

## PR Reflection Summary

### Refactoring
- [P1] <what was done>

### Tests
- [P2] <what was added>

### Documentation
- [P2] <what was updated>

### Cleanup
- [P1] <what was fixed>

### Deferred Items
- Fixed in PR: <what was addressed>
- Created #<num>: <title>
- (or: None identified)

(Use severity tags: P0, P1, P2. Omit P3 — see [review rubric](references/review-rubric.md).)

Guidelines

  • Pattern consistency is the highest-value check — a missed pattern update causes bugs across the codebase
  • Skip noise — see review rubric for severity calibration and what not to flag
  • Triage deferred items with the user — ask whether each item should be fixed now or deferred as a follow-up issue; only create issues for confirmed deferrals
  • Run quality gates before reporting — catch issues before the reviewer does
  • Prefer fresh context — a reviewer without implementation memory catches issues the author overlooks

Related Skills

After review: Use forge-address-pr-feedback to address reviewer feedback.

Example Usage

/forge-reflect-pr
/forge-reflect-pr -- pay extra attention to migration safety and missing regression tests

适合场景

01

用户想查找某类 Agent Skill 时

02

需要根据任务场景推荐可安装能力包时

03

需要对比不同来源的安装命令和来源信息时

能力概览

能力 1

按任务关键词查找相关 Skills

能力 2

展示可复制的安装命令

能力 3

保留来源站点、仓库和原始说明,方便继续核验

能力 4

展示第三方安全扫描或审计结果

安装后应在对应宿主中按原始 README 的触发条件使用;具体调用方式请以来源页面和 README 为准。

平台分布

Codex

38.73%
按下载量换算70

Claude

28.44%
按下载量换算51

Cursor

21.75%
按下载量换算39

Gemini CLI

8.85%
按下载量换算16

安全审计

Gen Agent Trust Hub

通过

Socket

通过

Snyk

通过

权限和风险

只读

该 Skill 主要提供规则、说明或参考内容,本身偏只读;真正读写文件、联网或执行命令仍取决于宿主 Agent 的任务。

安装前确认

本站仅展示第三方公开信息,不托管安装包,不提供自动安装或运行环境。安装前应自行审查源码、依赖和命令行为。当前只有一个来源,正式发布前建议补源仓库或其他目录站核验。

来源信息

继续浏览同类 Skills