third-party-code-vetting

Vet third-party code before it enters the repo or runs on a host — upstream patches, vendored scripts, installer snippets, peer-submitted cron scripts. Use when integrating external code of any kind.

インストールへ移動

リポジトリ: lukemcqueen/hermes-cortex
ソースの最終更新活動: 2026年8月5日 00:40
検出された SKILL.md の言語: 英語
スター: 4
フォーク: 1

インストール方法

デフォルトでは、最初にソースを確認する Prompt が選択されています。直接コマンドに切り替えるか、ローカルコピーをダウンロードすることもできます。

Codex または Claude でインストールこの Prompt をコピーして Codex、Claude、または他のアシスタントに貼り付けると、Skill ページを確認してインストールできます。

直接コマンドでは確認用 Prompt が省略されます。実行前にソースを確認してください。

npx skills add https://github.com/lukemcqueen/hermes-cortex --skill third-party-code-vetting

コマンドは1行のまま表示されます。コピー前に横へスクロールして全体を確認してください。

ローカルで確認しますか？SkillsMP が現在取得できるファイルをダウンロードできます。

ソースファイルを確認

インストールを決める前に、SKILL.md と SkillsMP に表示されている付属ファイルをお読みください。

SKILL.md

ソースの指示 · 読み取り専用プレビュー

検索に戻るインストールに戻る

agent-health-monitoring

Cross-server agent health monitoring using binary status vectors — deploy health endpoints on each agent, poll from orchestrator, alert on state transitions.

2026-09-01

langfuse-observability

Wire a self-hosted Langfuse instance to Hermes Agent — generate API keys, configure env vars, enable the bundled plugin, install SDK, and verify traces flow.

2026-09-01

enforcement-change-safety

Use before enforcement code changes or shared-repo commits.

2026-09-01

このリポジトリのすべてを見る

Third-Party Code Vetting

Vet external code before it enters the repo or runs on a host: upstream patches, vendored scripts, installer snippets, cron scripts from peers, MCP server code. The bar is provenance + behavior, not a glance at the diff.

Checklist

Source integrity — where did it come from? Commit hash, signed release, PR number, or named peer. Reject code with no provenance.

Read the whole file — never vet from a diff summary or a description. A 200-line upstream patch shipped on an unexamined premise is the classic failure (2026-08-03).

Static scan — adversarial-verify.py --file <script> --level A4 --gate on any scripts; secret-leak-detector.sh for PII/creds.

Behavior check — run it with boundary inputs in a sandbox: tmp workdir, no network, no sudo. Attack the implicit assumptions (Technique F), don't just execute the happy path.

Dependency check — stdlib-only unless declared; no silent network calls; no writes outside the declared scope.

Persistence check — what does it install or change? (crons, systemd units, config files, auto-start hooks). Reject silent persistence; require explicit opt-in.

License — compatible with the target repo (MIT for hermes-cortex).

Decision

Finding

Decision

No provenance

REJECT

Injection / exfiltration / hidden persistence

REJECT

Unaudited runtime network access

REJECT or fix + WARN

Provenance + clean scan + sandbox pass

APPROVE

Minor issues (paths, PII, naming)

APPROVE after fix, note it

Verify after approval

Commit with -F message citing the source (hash/PR/author)

Deploy and run the real invocation (doctor Script run evidence)

Never --no-verify to ship a vet-rejected change

Third-Party Code Vetting

Checklist

Source integrity — where did it come from? Commit hash, signed release, PR number, or named peer. Reject code with no provenance.

Read the whole file — never vet from a diff summary or a description. A 200-line upstream patch shipped on an unexamined premise is the classic failure (2026-08-03).

Static scan — adversarial-verify.py --file <script> --level A4 --gate on any scripts; secret-leak-detector.sh for PII/creds.

Behavior check — run it with boundary inputs in a sandbox: tmp workdir, no network, no sudo. Attack the implicit assumptions (Technique F), don't just execute the happy path.

Dependency check — stdlib-only unless declared; no silent network calls; no writes outside the declared scope.

Persistence check — what does it install or change? (crons, systemd units, config files, auto-start hooks). Reject silent persistence; require explicit opt-in.

License — compatible with the target repo (MIT for hermes-cortex).

Decision

Finding

Decision

No provenance

REJECT

Injection / exfiltration / hidden persistence

REJECT

Unaudited runtime network access

REJECT or fix + WARN

Provenance + clean scan + sandbox pass

APPROVE

Minor issues (paths, PII, naming)

APPROVE after fix, note it

Verify after approval

Commit with -F message citing the source (hash/PR/author)

Deploy and run the real invocation (doctor Script run evidence)

Never --no-verify to ship a vet-rejected change

name	third-party-code-vetting
version	1.0.0
description	Vet third-party code before it enters the repo or runs on a host — upstream patches, vendored scripts, installer snippets, peer-submitted cron scripts. Use when integrating external code of any kind.
triggers	["vet third-party code","external code review","vendored script","upstream patch"]

third-party-code-vetting

ソース情報

インストール方法

ソースファイルを確認

このリポジトリの他の Skills