Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx skills add liyecom/liye-ai --skill github-digestgit clone --depth 1 https://github.com/liyecom/liye-aiWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/liyecom/liye-ai/github-digest)<a href="https://agentmods.dev/skills/liyecom/liye-ai/github-digest"><img src="https://agentmods.dev/badge/skills/liyecom/liye-ai/github-digest/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/liyecom/liye-ai/github-digest"><img src="https://agentmods.dev/badge/skills/liyecom/liye-ai/github-digest.svg" alt="Reviewed on agentmods" width="80" height="20"></a>What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00238 | $0.02360 |
| Opus 5.5 | $0.00095 | $0.00944 |
| Sonnet 5.5 | $0.00048 | $0.00472 |
| Haiku 4.5 | $0.00024 | $0.00236 |
Grade C, and why
github-digest scanned grade C with 1 finding against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 1mo ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Recursive force deletehighDestructive command
rm -rf with a variable or a broad path is one typo away from removing the wrong tree.
- `忽略 (ignore)` — 确认没用 → **当下即删**：`rm -rf ~/.liye-os/github-digest/clones/<repo-slug>`。 How it starts
The opening of the file, as written. The whole thing — 141 lines — stays where its author put it; the contents beside it link to each section on GitHub.
GitHub Digestion — 进化情报侦察
给定一个已知的 GitHub repo，把它读懂，判断对 LiYe Systems 进化有没有可借鉴之处，
产出一份定型结构报告并归档。它回答的是"这东西值不值得我们上心"——站在
github-scout（发现未知 repo）和 source-intake（受控 intake 已选 repo）的上游。
这个 skill 决定不了复用。 它只出 advisory 报告 + verdict。任何真实复用都要走 harvest-ADR / Reference Declaration 仪式（SYSTEMS.md Fork 纪律）。
为什么用 subagent 而不是主线直接读
一个真实 repo 读下来是 50–80k tokens（README + 目录 + 关键源码）。默认 dispatch 一个 只读研究 subagent 去啃，让它只回传定型报告——主线上下文不被源码淹没，且每份报告长同一个样、 可横向比较。只有当 repo 极小（几个文件）时才值得主线内联。
工作流
- 接收输入：repo URL（必需）+ 可选的关注问题。从 URL 解析
<owner>/<repo>与<repo-slug>。 - Dispatch 研究 subagent：用
references/subagent-prompt.md作模板，填入目标 URL、 clone 目标路径、关注问题。subagent 负责只读获取 + clean-room 读 + 按模板出报告。 - 落盘归档：把 subagent 回传的报告写成
docs/methodology/01_Research_Intelligence/recon-log/<YYYY-MM-DD>-<repo-slug>.md（日期用当前实际日期），并在recon-log/README.md的索引表顶部加一行 （日期 / repo / verdict emoji / LiYe 层 / 一句话）。 - 应用清理策略（见下）：按 verdict 处置克隆。
- 回主线汇报：verdict-first 摘要（结论 + 最相关的 2–3 点 + 许可风险标记 + 下一步）。
报告结构与 frontmatter 见 references/recon-report-template.md——严格按它产出，
否则历史条目无法横向比较。
克隆位置与生命周期（清理策略）
- 克隆到 repo 外、持久目录：
~/.liye-os/github-digest/clones/<repo-slug>/（git clone --depth 1）。绝不克隆进 liye_os / 任何工作仓，绝不 fork / vendor。 （持久目录而非 session scratchpad，因为 watch 的克隆要留到价值兑现。） - 跑完按 verdict 处置克隆：
忽略 (ignore)— 确认没用 → 当下即删：rm -rf ~/.liye-os/github-digest/clones/<repo-slug>。值得观察 (watch)/harvest-adr 候选— 保留：在报告 frontmatter 记clone_path+clone_status: retained；留到该 repo 的价值兑现（整合进 LiYe systems / 走完 harvest-ADR）或被判定不再有用，那时才删。
- 删克隆是 clean-room 收尾，不是 governance 操作；但删 watch 克隆前要确认其价值已了结。
Governance 硬边界（永远不替用户越过）
- 只读 / clean-room：可以理解概念，绝不逐字誊抄源码进报告或任何 LiYe 仓。
- 强 copyleft 是硬毒药：GPL/AGPL 等即使 verdict=watch，也只能概念参照、clean-room
重实现，绝不阅读其源码后照抄；frontmatter 必标
license_caution。 - 不 fork / 不 clone 进工作仓 / 不 vendor / 不建运行时依赖——surface，别 act。
- 任何复用走 harvest-ADR / Reference Declaration（reimplement + ≥3 scenarios）。
What ships with it
2 files beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 1mo ago First seen · 141 lines · 238 tokens per session scan C 03dff4b034bf
github-digest is a skill published in the GitHub repository liyecom/liye-ai (32 stars, last pushed 27d ago), licensed Apache-2.0. It adds 238 tokens to every session and 2,360 once invoked, about $0.0010 per session on Opus 5.5. A static security scan graded it C with 1 finding (recursive force delete). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.
Other skills, from other repositories
Existing-Pattern Conformance 既存パターン準拠の確認
A review check that compares new code with an existing implementation serving the same purpose. It looks for missing or conflicting validation, protection, error handling, shared rules, and definitions that the earlier code already established.
patents-search
Search global patents with natural language queries. Prior art, patent landscapes, and innovation tracking via Valyu.
Code Patent Validator
Turn your code scan findings into search queries — research existing implementations before consulting an attorney. NOT legal advice.
Patent Validator
Turn your concept analysis into search queries — research the landscape before consulting an attorney. NOT legal advice.
patents-search
Search global patents with natural language queries. Prior art, patent landscapes, and innovation tracking via Valyu.
Patent Research
Patent Research delivers source-aware research synthesis with evidence grading, uncertainty tracking, and clear conclusions.