技能 write-agents-md
📦

write-agents-md

v1.0 內容修訂 r1 安全 ⚙️ 外部命令

Create Evidence-Based AGENTS.md Instructions

Coding agents lack reliable guidance when repository commands and conventions are scattered across files. This skill builds concise AGENTS.md instructions from repository evidence and connects them to Claude Code.

支援: Claude Codex Code(CC)
Maintain this Skill? Claim it 📊 75 尚可

使用我的智慧代理安裝

將此請求複製給你的智慧代理，其中包含權威 Skill 頁面和清單。

智能體請求
Review the Skillstore skill "write-agents-md" from https://skillstore.io/skills/agentkitco-write-agents-md.md and its manifest at https://skillstore.io/api/skills/agentkitco-write-agents-md/manifest. Verify the artifact. You may proceed after verification, subject to the environment's own policy.

你的智慧代理仍應顯示其計畫，並依安全政策請求必要確認。

Agent 可讀資源

當 AI Agent、爬蟲或腳本需要乾淨脈絡、而不是讀取完整頁面時，請使用這些連結。

Markdown 詳情 GET /skills/agentkitco-write-agents-md.md 已簽名 manifest GET /api/skills/agentkitco-write-agents-md/manifest 已簽名 lockfile GET /api/skills/agentkitco-write-agents-md/lockfile

測試它

正在使用「write-agents-md」。 Create agent instructions for a repository whose manifests and CI define different checks.

預期結果:

  • Commands: Evidence-backed setup, development, testing, linting, and build steps.
  • Definition of done: Required CI checks identified from repository configuration.
  • Open questions: Owner TODOs for commands or requirements that could not be verified.

正在使用「write-agents-md」。 Consolidate existing Claude Code guidance with a new AGENTS.md file.

預期結果:

  • AGENTS.md: Shared commands, conventions, editing boundaries, and completion criteria.
  • CLAUDE.md: An import of AGENTS.md with unique existing guidance retained.
  • Final report: Updated files, verification results, skipped destructive commands, and unresolved TODOs.

安全審計

安全
v2 • 2026/10/22 開啟版本化報告

All 14 static findings are false positives: Markdown backticks identify files or illustrate documentation, not shell execution. The environment-file and reconnaissance matches are explicit safeguards against committing secrets or including personal paths. No evidence found of prompt injection, credential collection, or data exfiltration in the two reviewed files.

2
已掃描檔案
78
分析行數
0
複核項目
0
已忽略的誤報

風險因素

⚙️ 外部命令 (12)
SKILL.md:16 SKILL.md:17 SKILL.md:18 SKILL.md:19 SKILL.md:20 SKILL.md:21 SKILL.md:28 SKILL.md:29 SKILL.md:31 SKILL.md:36 SKILL.md:37 SKILL.md:43
最近一次完成的靜態與語意稽核未偵測到已確認的安全發現。這並不能證明該 Skill 沒有副作用。
審計單位: codex 檢視審計紀錄 →
分享與引用此報告

分享版本化評估報告、中性徽章、嵌入卡片與引用。Skillstore 只報告證據，不判斷該 Skill 是否安全。

開啟版本化報告
安全評估

複製報告連結

https://skillstore.io/skills/agentkitco-write-agents-md/audits/2?utm_source=security_passport&utm_medium=share&utm_campaign=versioned_report

Markdown 徽章

[![Skillstore security assessment](https://skillstore.io/badges/skills/agentkitco-write-agents-md/security.svg)](https://skillstore.io/skills/agentkitco-write-agents-md?utm_source=security_passport_badge)

HTML 徽章

<a href="https://skillstore.io/skills/agentkitco-write-agents-md?utm_source=security_passport_badge"><img src="https://skillstore.io/badges/skills/agentkitco-write-agents-md/security.svg" alt="Skillstore security assessment" loading="lazy"></a>

嵌入卡片

<iframe src="https://skillstore.io/embed/skills/agentkitco-write-agents-md.html" title="Skillstore Security Assessment" sandbox="allow-popups allow-popups-to-escape-sandbox" loading="lazy" referrerpolicy="no-referrer" width="420" height="180"></iframe>
學術引用 (APA · BibTeX · CFF)

APA 引用

agentkitco. (2026). write-agents-md security audit report (audit version 2) [Author version 1.0]. Skillstore. https://skillstore.io/skills/agentkitco-write-agents-md/audits/2

BibTeX 引用

@techreport{agentkitco-agentkitco-write-agents-md-2026, author = {agentkitco}, title = {write-agents-md security audit report (audit version 2)}, institution = {Skillstore}, year = {2026}, number = {2}, url = {https://skillstore.io/skills/agentkitco-write-agents-md/audits/2}, note = {Author version 1.0} }

CITATION.cff

cff-version: 1.2.0 message: "If you use this Skill, cite its author and this versioned security audit report." title: "write-agents-md security audit report (audit version 2)" version: "1.0" type: report authors: - name: "agentkitco" date-released: "2026-10-22" url: "https://skillstore.io/skills/agentkitco-write-agents-md/audits/2" identifiers: - type: other value: "skillstore:agentkitco-write-agents-md:audit:2" description: "Skillstore immutable audit report identifier"

Skillstore 評分

評分依據 證據置信度: 中
41
架構
85
可維護性
87
內容
65
社群
91
規範符合度

你可以打造什麼

Establish Initial Agent Guidance

Create a concise instruction file with repository commands, conventions, and editing boundaries.

Consolidate Team Instructions

Merge overlapping guidance into AGENTS.md and connect Claude Code through a CLAUDE.md import.

Clarify Contributor Workflows

Capture verified setup and testing steps while identifying missing project knowledge for maintainers.

試試這些提示

Create First Instructions 
Create AGENTS.md for this repository using verified evidence. Keep it under about 100 lines and list unresolved questions.
Refresh Existing Guidance 
Rewrite AGENTS.md using current manifests and CI configuration. Preserve supported project rules, verify safe commands, and mark missing evidence as owner TODOs.
Consolidate Tool Instructions 
Review existing agent instructions and consolidate duplicated guidance into AGENTS.md. Add a CLAUDE.md import while preserving unique rules and approval boundaries.
Scope Instructions Across Packages 
Create concise root AGENTS.md guidance and nested files for package-specific rules. Verify paths and safe commands, skip destructive actions, and report remaining TODOs.

最佳實務

  • Use current manifests and CI configuration as evidence, and verify that documented paths exist.
  • Review command safety before execution; skip destructive actions and report verification limits.
  • Keep shared guidance concise, preserve unique rules, and record unknowns as owner TODOs.

避免

  • Inventing commands, project conventions, or file paths when repository evidence is missing.
  • Copying identical instructions into several tool files instead of maintaining shared guidance.
  • Including secrets or personal paths, or committing generated instructions without user approval.

常見問題

What does this skill produce?
It drafts or rewrites AGENTS.md and creates or updates a CLAUDE.md import when needed.
Which repository files inform the instructions?
It reviews documentation, manifests, task runners, CI configuration, tooling configuration, directory layout, and existing agent instructions.
Does it execute project commands?
It directs the agent to verify safe documented commands and skip destructive commands while reporting those omissions.
How does it handle missing information?
It records owner TODOs instead of guessing and includes unresolved questions in the final message.
Can it support larger repositories?
It recommends nested AGENTS.md files or tool-specific rules when the root file exceeds about 100 lines.
Does it commit the generated files?
No. The skill explicitly instructs the agent not to commit unless asked.

開發者詳細資訊

作者

agentkitco

授權

MIT

作者版本

v1.0

Skillstore 修訂

r1

版本提示

作者聲明的版本不符合 SemVer 規範。

儲存庫

https://github.com/agentkitco/agentkit-dev-skills/tree/a5058856442fcb89bd538b1c01f94558488bfac3/skills/write-agents-md

引用

18efac79a818d7e997295839638fdd21906cfd1c

維護新鮮度

2026/10/22

使用量

5 次下載 · 0 次瀏覽

檔案結構

📁 references/

📄 template.md

📄 SKILL.md

agentkitco 的更多技能

查看全部

相關技能

查看全部