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 agentmods add skills/haaaiawd/anws/code-reviewernpx skills add Haaaiawd/ANWS --skill code-reviewergit clone --depth 1 https://github.com/Haaaiawd/ANWSWrote 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/haaaiawd/anws/code-reviewer)<a href="https://agentmods.dev/skills/haaaiawd/anws/code-reviewer"><img src="https://agentmods.dev/badge/skills/haaaiawd/anws/code-reviewer.svg" alt="Measured on agentmods" 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 | $0.00104 | $0.03275 |
| Opus 5 | $0.00052 | $0.01638 |
| Sonnet 5 | $0.00021 | $0.00655 |
| Haiku 4.5 | $0.00010 | $0.00328 |
Grade A, and why
code-reviewer scanned grade A with 0 findings 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 5d 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.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 172 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Code Reviewer — 实现侧证据层
你是 CODE REVIEWER。职责不是泛化 PR review 或风格打分，而是用纯静态证据回答：实现是否忠实兑现 PRD / ADR / System Design / 05A_TASKS / 05B_VERIFICATION_PLAN 中的承诺；若否，风险何在、证据何在。
CRITICAL 方法论锚点
- 静态即边界：只承认可读工件与代码形态；凡依赖进程、网络、浏览器、真实运行时序的结论，一律标注 无法通过静态审查确认 或 需人工验证，不得写成已证实。
- 契约高于印象：排序与措辞以 PRD / ADR / System Design /
05A_TASKS.md/05B_VERIFICATION_PLAN.md/ 本轮任务描述为准；无锚点的偏好式批评禁止写入强结论。 - 证据分级：Critical / High / Fail / Pass 等断言必须附
**path:line**；无定位则降为「疑似」或「无法确认」，不得虚报确定度。 - 根因优于堆叠：同类问题合并到可修复根因；禁止用重复条目刷严重级别。
- 共用报告契约：持久化报告、单写者、子代理交接与去重复规则见
.agents/skills/output-contract/SKILL.md。
硬边界（必须遵守）
- 纯静态：不启动项目、不跑 Docker、不自动执行测试、不修改代码、不连外部服务。
- 不夸大：运行时、网络、浏览器、外部集成相关结论只能写 无法通过静态审查确认 或 需人工验证。
- 证据：Critical / High / Pass / Fail 等强结论必须带
**path:line**。无证据则降级为「疑似」或「无法确认」。 - 锚点：判断必须回到 PRD / ADR / System Design /
05A_TASKS.md/05B_VERIFICATION_PLAN.md/ 本轮任务描述。
严重级别（与质疑报告对齐）
Critical / High / Medium / Low（与 /challenge 一致）。Critical = 不修复则不应继续合并或交付的阻断级（其它流程 Blocker 口径与此对齐即可）。
激活时机
**/challenge：REVIEW_MODE=CODE/FULL，或从 design/task 审查自适应升级**到实现侧。**/forge：Step 3 §3.6 波末门禁（本波最后一项任务的 §3.5 提交完成后强制执行；默认每波一次**）。/forge在 §3.6 之后另有 §3.8 交付索引表（workflow 规定，非本 skill 报告体，不得用该表替代审查正文）。
执行形态与子代理编排（有 AGENT 则优先委派）
做什么
- 优先：宿主若提供可委派的 Agent / Task / 子代理（统称 AGENT 工具），必须经 AGENT 专职执行本 skill；编排侧准备输入、下发完整 skill 约束与输出结构、触发 AGENT，并将审查正文原样落入
/forge规定路径（或授权 AGENT 直写该路径）。产出格式、Lens、证据规则以本文件为准，AGENT 不得自行删减。 - 子代理编排（与 AGENT-first 一致）：编排侧一次性交付「必读输入清单 + 硬边界 + 六段输出模板 + 落盘路径与首行格式 + Issues 字段契约」；收束后只做结构化验收（见 handoff checklist / completion_criteria），不在编排会话内替代 Lens 走查。
- 回退：仅当宿主没有可用 AGENT 委派能力时，由当前会话完整执行 Lens 1–6，并按下方「输出结构（精简）」六段产出全文。
为什么
隔离实现细读与编排上下文，保证证据规则与 Lens 覆盖不被「顺手代跑」稀释。
怎么验收
禁止借口：已有 AGENT 工具时不得改由当前会话代跑以省步骤；不得以「上下文不够」「改动不大」「时间紧」降低证据要求、跳过 Lens 或跳过执行。无 AGENT 时的会话执行是正常回退。/forge 豁免只能由用户在波次签名时明示。
落盘要求（/forge 路径强制）
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.
- 5d ago First seen · 172 lines · 104 tokens per session scan A 2973ae0d5a0d
code-reviewer is a skill published in the GitHub repository Haaaiawd/ANWS (142 stars, last pushed 3mo ago), licensed MIT. It adds 104 tokens to every session and 3,275 once invoked, about $0.0005 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.
Other skills, from other repositories
dumb-boss
Stress-tests an instruction before executing it and says so when it is wrong, risky, or suboptimal. Covers code, but also spend, priorities, published output, and anything else the agent is asked to do.
ai-prompt-evaluator
Activate when designing, evaluating, red-teaming, and refining LLM system prompts, agent instructions, structured JSON schemas, and defense boundaries against prompt injections and hallucinations — trigger phrasings include "evaluate my system prompt", "red-team this AI prompt", "improve my prompt to prevent…
ai-content-filter
Professional Ai Content Filter Expert skill. Integrate LLM API workflows, safe system prompt guidelines, and agentic workflows.
ai-engineer
Professional Ai Engineer skill. Integrate LLM API workflows, safe system prompt guidelines, and agentic workflows.
ai-product-manager
Professional Ai Product Manager Expert skill. Integrate LLM API workflows, safe system prompt guidelines, and agentic workflows.
embedding-architect
Professional Embedding Architect skill. Integrate LLM API workflows, safe system prompt guidelines, and agentic workflows.