Token导航 LogoToken导航

Agent Skill · 研究检索

create-review(创建评论)

create-review 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。

create-review 是研究检索类 Agent Skill,可用宿主为 Codex、Claude、Cursor、Gemini CLI,通过GitHub 仓库获取,权限为只读;许可证信息还不明确,第三方扫描有提醒项,按安装量在研究检索类的 73,675 个 Skill 里排第 61,129。

总安装 356周安装 15GitHub Stars 5下载 125核验 2026-05-01

安装说明

本站只整理中文说明,不托管安装包,也不代用户安装。
GitHub
许可证
unknown

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:create-review(创建评论)
来源仓库:https://github.com/generaljerel/chalk-skills
仓库路径:skills/create-review
安装命令:
npx skills add https://github.com/generaljerel/chalk-skills --skill create-review
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/generaljerel/chalk-skills --skill create-review

这个 Skill 是做什么的

研究检索类 Skill 主要解决「找资料、核对事实、把来源整理清楚」的问题,通常会让助手先检索再总结。

启动评审流程并为任意 AI 审阅者生成提示词。适用于多模型对比评审和会话化管理场景。create-review 属于研究检索类 Skill,可作为该场景下的辅助能力补充。自动检测 reviewer 和 session,生成 paste-ready 提示。

安装需确认 .chalk/reviews/ 目录结构和分支命名规则。支持跨会话持久化评审上下文,提升反馈连续性。

来源仓库内路径:skills/create-review

上手步骤

4 步
  1. 1

    确认自己的宿主工具

    公开资料标注支持 Codex、Claude、Cursor、Gemini CLI;先确认自己用的是其中之一,再按对应工具的 Skills 目录放置。

  2. 2

    拿到安装命令

    可以直接复制的命令是 npx skills add https://github.com/generaljerel/chalk-skills --skill create-revie…;命令里的占位符换成自己的路径或密钥再执行。

  3. 3

    安装后重启宿主

    命令执行完要完全退出宿主工具再打开,让新装的 Skill 被重新扫描到,否则列表里可能看不到。

  4. 4

    按最小权限试用

    它标注为只读,先用只读任务试一次,确认行为符合预期再扩大使用范围。

常见问题

6 条

create-review 怎么安装?

在终端执行 npx skills add https://github.com/generaljerel/chalk-skills --skill create-review。装完完全退出宿主工具再打开,让它重新扫描本地 Skill。

create-review 支持哪些 AI 工具?

公开资料标注支持 Codex、Claude、Cursor、Gemini CLI。其他工具只要能读取本地 Skill 目录,一般也能手动放进去,但要自己确认触发方式是否一致。

create-review 需要什么权限?

权限级别是只读。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。

create-review 是免费的吗?

公开资料没有标注明确许可证,收费和商用限制都无法确认。正式使用前请到来源仓库核对授权说明。

create-review 用的人多吗?

公开数据是累计安装 356 次,近一周安装 15 次,下载 125 次,GitHub 5 星。这些数字来自第三方目录统计,只能当作热度参考,不代表质量或安全性。

create-review 安全吗?

第三方扫描有提醒项,说明存在需要人工确认的行为。安装前请打开源码看清它读了什么、联网到哪里。

值得选的地方和要先想清楚的地方

值得选的地方

  • 同时适配 4 种工具,换客户端不用重新找替代品。
  • 标注为只读权限,试用阶段不容易误改本地文件。

先想清楚的地方

  • 许可证没有明确标注,能不能商用、能不能再分发都不确定。
  • 第三方扫描存在提醒项或未通过项,需要人工逐条确认。
  • 第三方目录整理,版本更新和行为细节以来源仓库为准。

适不适合你

更适合

  • 已经在用 Codex,想直接扩展它能力的人。
  • 需要处理研究检索类任务、又不想每次都手写提示词的人。
  • 只想先试试效果、不希望它改动本地文件的人。

不太适合

  • 需要明确授权才能用到公司项目里的场景。
  • 需要厂商级技术支持和 SLA 的生产系统。

研究检索类 Skill 对比

6 个同类
Skill总安装GitHub 星权限许可证扫描
find-skills

寻找技能 · Codex、Claude

31,199,81316,486操作浏览器MIT有提醒
self-improving-agent

自我改进 Agent · OpenClaw

16,464,0023,396接触敏感数据MIT-0通过
soultrace

灵魂追踪 · Codex、Claude

7,588,94412需要联网未明确有提醒
self-improving

自我完善 · OpenClaw

6,829,2841,046需要联网MIT-0通过
microsoft-foundry

微软代工厂 · Codex、Claude

6,614,144764调用外部服务未明确有提醒
create-review

创建评论 · 本页

3565只读未明确有提醒

按安装量取同分类前几名;数字来自第三方目录统计,只作热度参考,不代表质量结论。

SKILL.md 原文

Create Review

Bootstrap the review pipeline and generate a paste-ready review prompt for any AI reviewer.

Step 1: Determine the reviewer and session

Reviewer: If the user provided $ARGUMENTS, sanitize it to a safe kebab-case string (lowercase, strip any characters that aren't alphanumeric or hyphens, collapse multiple hyphens) and use that as the reviewer name (e.g. codex, gemini, gpt4, claude). If no argument, use generic.

Session: Detect from context:

  1. If .chalk/reviews/ exists, check for the most recent session directory
  2. Otherwise, infer from the current branch name (kebab-case)
  3. If on main/master, ask the user

Store as {reviewer} and {session}.

Step 2: Bootstrap the review pipeline

Check if .chalk/reviews/scripts/pack.sh exists. If not, bootstrap the full pipeline:

mkdir -p .chalk/reviews/scripts .chalk/reviews/templates .chalk/reviews/sessions

Create .chalk/reviews/scripts/pack.sh

This script generates a review context pack from git state:

#!/usr/bin/env bash
set -euo pipefail

BASE_REF="${1:-origin/main}"
SESSION="${2:-adhoc}"
OUTPUT_PATH="${3:-.chalk/reviews/sessions/${SESSION}/pack.md}"

# Resolve base ref
if ! git rev-parse --verify "$BASE_REF" >/dev/null 2>&1; then
  for candidate in main origin/main master origin/master; do
    if git rev-parse --verify "$candidate" >/dev/null 2>&1; then
      BASE_REF="$candidate"
      break
    fi
  done
fi

MERGE_BASE="$(git merge-base HEAD "$BASE_REF" 2>/dev/null || echo "")"
if [ -z "$MERGE_BASE" ]; then
  MERGE_BASE="$(git rev-list --max-parents=0 HEAD | tail -n 1)"
fi

mkdir -p "$(dirname "$OUTPUT_PATH")"

{
  echo "# Review Pack"
  echo
  echo "- Session: \`$SESSION\`"
  echo "- Generated: $(date -u +"%Y-%m-%d %H:%M:%S UTC")"
  echo "- Base ref: \`$BASE_REF\`"
  echo "- Merge base: \`${MERGE_BASE:0:12}\`"
  echo "- Head: \`$(git rev-parse --short HEAD)\`"
  echo
  echo "## Diff Stat"
  echo '```'
  git diff --stat "$MERGE_BASE"..HEAD 2>/dev/null || echo "(no committed diff)"
  echo '```'
  echo
  echo "## Changed Files"
  CHANGED="$(git diff --name-only "$MERGE_BASE"..HEAD 2>/dev/null || true)"
  if [ -n "$CHANGED" ]; then
    echo "$CHANGED" | while IFS= read -r f; do [ -n "$f" ] && echo "- $f"; done
  else
    echo "- (none)"
  fi
  echo
  echo "## Commit Log"
  echo '```'
  git log --oneline "$MERGE_BASE"..HEAD 2>/dev/null || echo "(no commits ahead of base)"
  echo '```'
  echo
  echo "## Working Tree Status"
  echo '```'
  git status --short
  echo '```'
} > "$OUTPUT_PATH"

echo "PACK_PATH=$OUTPUT_PATH"

Create .chalk/reviews/scripts/render-prompt.sh

This script combines pack + handoff + reviewer template into a prompt:

#!/usr/bin/env bash
set -euo pipefail

REVIEWER="${1:?Usage: render-prompt.sh <reviewer> [pack-path] [handoff-path] [output-path]}"
PACK_PATH="${2:-}"
HANDOFF_PATH="${3:-}"
OUTPUT_PATH="${4:-}"
SESSION="${5:-}"
SCRIPT_DIR="$(cd "$(dirname "$0")" && pwd)"
ROOT_DIR="$(dirname "$SCRIPT_DIR")"

REVIEWER_TITLE="$(echo "$REVIEWER" | awk '{print toupper(substr($0,1,1)) substr($0,2)}')"

if [ -z "$SESSION" ] && [ -f "$ROOT_DIR/.current-session" ]; then
  SESSION="$(cat "$ROOT_DIR/.current-session")"
fi
SESSION="${SESSION:-adhoc}"

[ -z "$PACK_PATH" ] && PACK_PATH="$ROOT_DIR/sessions/$SESSION/pack.md"
[ -z "$HANDOFF_PATH" ] && HANDOFF_PATH="$ROOT_DIR/sessions/$SESSION/handoff.md"
[ -z "$OUTPUT_PATH" ] && OUTPUT_PATH="$ROOT_DIR/sessions/$SESSION/${REVIEWER}.prompt.md"

if [ ! -f "$PACK_PATH" ]; then
  echo "Pack not found at $PACK_PATH. Run pack.sh first." >&2
  exit 1
fi

mkdir -p "$(dirname "$OUTPUT_PATH")"

{
  echo "# $REVIEWER_TITLE Review Request"
  echo

  # Use the universal reviewer template
  TEMPLATE="$ROOT_DIR/templates/reviewer.template.md"
  if [ -f "$TEMPLATE" ]; then
    cat "$TEMPLATE"
  fi

  echo
  echo "---"
  echo
  echo "## Review Pack"
  cat "$PACK_PATH"

  if [ -f "$HANDOFF_PATH" ]; then
    echo
    echo "---"
    echo
    echo "## Handoff"
    cat "$HANDOFF_PATH"
  fi
} > "$OUTPUT_PATH"

echo "PROMPT_PATH=$OUTPUT_PATH"

Create .chalk/reviews/scripts/copy-prompt.sh

#!/usr/bin/env bash
set -euo pipefail

REVIEWER="${1:?Usage: copy-prompt.sh <reviewer> [pack] [handoff] [output] [session]}"
SCRIPT_DIR="$(cd "$(dirname "$0")" && pwd)"

OUTPUT="$(bash "$SCRIPT_DIR/render-prompt.sh" "$@")"
echo "$OUTPUT"

PROMPT_PATH="$(echo "$OUTPUT" | sed -n 's/^PROMPT_PATH=//p' | head -1)"
if [ -z "$PROMPT_PATH" ] || [ ! -f "$PROMPT_PATH" ]; then
  echo "Could not resolve prompt path." >&2
  exit 1
fi

COPIED=0
if command -v pbcopy >/dev/null 2>&1; then
  pbcopy < "$PROMPT_PATH" && COPIED=1 && echo "CLIPBOARD=pbcopy"
elif command -v xclip >/dev/null 2>&1; then
  xclip -selection clipboard < "$PROMPT_PATH" && COPIED=1 && echo "CLIPBOARD=xclip"
elif command -v wl-copy >/dev/null 2>&1; then
  wl-copy < "$PROMPT_PATH" && COPIED=1 && echo "CLIPBOARD=wl-copy"
fi

if [ "$COPIED" -eq 0 ]; then
  echo "CLIPBOARD=none (copy manually from $PROMPT_PATH)"
fi

Create .chalk/reviews/templates/reviewer.template.md

Only create if it does not already exist (preserve user customizations):

You are acting as an independent code reviewer.

Primary objective:
- Find real defects and risks in changed lines only.
- Prioritize actionable, high-signal output over style commentary.
- Report defects and risks, not style preferences.

Output format (required):

1. `## Verdict`
   - `Block merge: yes|no`
   - `Blocking findings: P0=<n>, P1=<n>`
   - If no P0/P1 findings, include exact text: `No blocking findings`.

2. `## Findings`
   - Use a markdown table with columns:
     - `ID` (R-001, R-002, ...)
     - `Severity` (P0 = critical | P1 = high | P2 = medium | P3 = low)
     - `Category` (Security | Correctness | Performance | Reliability | Testing)
     - `File:Line`
     - `Issue` — concise summary
     - `Failure mode` — what breaks and when
     - `Suggested fix` — actionable next step
     - `Confidence` (0.00–1.00)

3. `## Testing Gaps`
   - List missing tests that could hide regressions.

4. `## Open Questions`
   - Only unresolved assumptions that affect correctness.

Rules:
- Review changed lines only.
- Focus on correctness, security, reliability, and regression risk.
- Do not comment on formatting, import ordering, or trivial naming.
- Do not suggest broad refactors unless required for correctness.
- Keep recommendations patch-oriented and specific to the failure mode.
- If no blocking issues exist, explicitly state: `No blocking findings`.

Make scripts executable

chmod +x .chalk/reviews/scripts/pack.sh .chalk/reviews/scripts/render-prompt.sh .chalk/reviews/scripts/copy-prompt.sh

Create .chalk/reviews/PIPELINE.md

Write a brief usage guide explaining the pipeline, available scripts, and how to add custom reviewer templates. Refresh this on every run.

Step 3: Resolve the base branch

  1. git merge-base main HEAD → if it works, use it
  2. Try origin/main, then master, origin/master
  3. Store as {base}

Step 4: Check for a handoff

Look for .chalk/reviews/sessions/{session}/handoff.md. If it exists, it will be included in the prompt. If not, warn the user that no handoff was found and suggest running /create-handoff first, but continue anyway.

Step 5: Generate the review pack

bash .chalk/reviews/scripts/pack.sh "{base}" "{session}" ".chalk/reviews/sessions/{session}/pack.md"

Step 6: Generate the review prompt

bash .chalk/reviews/scripts/render-prompt.sh "{reviewer}" \
  ".chalk/reviews/sessions/{session}/pack.md" \
  ".chalk/reviews/sessions/{session}/handoff.md" \
  ".chalk/reviews/sessions/{session}/{reviewer}.prompt.md" \
  "{session}"

Step 7: Copy to clipboard

bash .chalk/reviews/scripts/copy-prompt.sh "{reviewer}" \
  ".chalk/reviews/sessions/{session}/pack.md" \
  ".chalk/reviews/sessions/{session}/handoff.md" \
  ".chalk/reviews/sessions/{session}/{reviewer}.prompt.md" \
  "{session}"

Step 8: Report to the user

Show:

  • The prompt file path
  • Whether it was copied to clipboard
  • Suggest: paste the prompt into any AI model (Codex, Gemini, GPT, Claude, etc.)
  • To run multiple reviews: run the skill again with a different reviewer name for labeling (e.g. /create-review gemini)

Also mention:

  • The reviewer template at .chalk/reviews/templates/reviewer.template.md can be customized
  • Each run with a different reviewer name creates a separate prompt file in the session directory

Step 9: Save current session

Write the session name to .chalk/reviews/.current-session so subsequent runs can pick it up.

Rules

  • Only create template files if they don't already exist — preserve user customizations
  • Always refresh scripts (pack.sh, render-prompt.sh, copy-prompt.sh) to latest version
  • Always refresh PIPELINE.md to latest version
  • Do NOT modify any source code
  • If no git changes exist ahead of base, warn the user but still generate (they may have local uncommitted work)

平台分布

Codex

34.65%
按下载量换算43

Claude

31.84%
按下载量换算40

Cursor

20.44%
按下载量换算26

Gemini CLI

9.7%
按下载量换算12

安全审计

Gen Agent Trust Hub

可疑

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。来源安全扫描存在提醒或未通过项,安装前请自行复核。目前只有一个公开来源,建议同时核对源仓库。

继续浏览同类 Skills