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 tranfu-labs/tranfu-skills --skill credibility-reviewgit clone --depth 1 https://github.com/tranfu-labs/tranfu-skillsWrote 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/tranfu-labs/tranfu-skills/credibility-review)<a href="https://agentmods.dev/skills/tranfu-labs/tranfu-skills/credibility-review"><img src="https://agentmods.dev/badge/skills/tranfu-labs/tranfu-skills/credibility-review/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/tranfu-labs/tranfu-skills/credibility-review"><img src="https://agentmods.dev/badge/skills/tranfu-labs/tranfu-skills/credibility-review.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 | $0.00201 | $0.07548 |
| Opus 5 | $0.00101 | $0.03774 |
| Sonnet 5 | $0.00040 | $0.01510 |
| Haiku 4.5 | $0.00020 | $0.00755 |
Grade A, and why
credibility-review 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 12d 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 — 461 lines — stays where its author put it; the contents beside it link to each section on GitHub.
credibility-review
用两个独立 agent 同时审稿:A 轨模拟读者(双轴 耐心 / 信任 标量),B 轨反模式 catalog(语言层枚举)。两个独立 fresh agent 并行——A 不开规则书素读,B 跑 catalog;最后三态合判:两轨都 PASS = 可发;都 FAIL = 退稿;一 PASS 一 FAIL = 待审查。只诊断,不重写。规则源永远是
tranfu-site/goal-docs/05-design-踩坑记-final.md与06-design-养成记-final.md。
为什么双轨:B 轨能抓"语言层有违规"(PR 套话 / 教程标题 / 隐蔽 hero);但 B 轨 0 命中不等于文章有内容——一个 fluent 的 PR 写手能写一篇全部避开反模式但全空的文章。A 轨抓的就是这种"语言干净但读起来空"——通过模拟读者的耐心和信任在读的过程中怎么消耗。两件事互补,缺一不可。
§0 兼容性合同
本 skill 假设:
- BSD grep(macOS 默认)或 GNU grep ≥ 2.6
- 不依赖 PCRE / -P / \b / \d / \w
- 中文 UTF-8 默认 work,无需 LANG / LC_ALL
- 待审稿件用原始路径 <target>,不复制到 /tmp
- 中英双语:英文用 `grep -nEi`(B 轨)
修订 grep 时必须在 macOS 默认 zsh + /usr/bin/grep 下实测。
§1 何时调用 / 不调用
调用:用户给一段 markdown 或一个 .md 路径,说"看下能不能发 / 帮我审一下 / 抓一下问题 / 检查文章";用户在 tranfu-site/src/content/posts/*.md 上停下来求审;用户贴一篇 AI 行业文章问"这写得怎么样 / 像不像 PR 稿"。
不调用:用户要重写 / 润色 / 扩写 / 打分 / 评级;用户要"全文点评 / 建议从几个维度展开"。
§2 工作流
Step 1 主 agent 识别体裁(踩坑记 / 养成记 / 其它)
Step 2 主 agent 用 Agent 工具并发起两个 fresh sonnet sub-agent:
- A 轨 (sub-agent A):读 §3 + 文章, 输出读者轨迹 + verdict
- B 轨 (sub-agent B):读 §4 + 文章, 输出 catalog 命中清单 + verdict
两个 sub-agent **互不可见对方的 prompt 和 output**——必须独立判断。
Step 3 两个 sub-agent 都返回后, 主 agent 按 §5 三态合判
Step 4 主 agent 按 §6 schema 汇总报告
硬规则:A 轨 sub-agent 的 prompt 不得包含 B 轨内容(不让它知道反模式 catalog),否则 A 会污染成 B 的子集,失去独立判断价值。两轨必须真独立。
任一步发现退稿级违规也不要提前终止——一篇文章常同时踩多个坑,一次性把全部问题报给作者比挤牙膏式反馈更有用。
Step 1 — 识别体裁(主 agent 同步执行)
按以下顺序判断:
- 看 frontmatter
tag:踩坑→ 踩坑记;养成→ 养成记。 - 无 frontmatter:看时间跨度——单点事件(小时 / 几天)= 踩坑记;≥ 2 个月演化 = 养成记。
- 仍无法判断 → 直接问用户"这篇你打算发到哪个栏目",不要猜。
非踩坑/养成的文章(资讯爆料 / PR 通稿 / research essay / advice tutorial / postmortem 等):体裁 = "其它"。A 轨用通用读者预期 set(见 §3.4 末),B 轨跳过结构核对仍跑反模式 catalog。
规则源:踩坑记 = tranfu-site/goal-docs/05-design-踩坑记-final.md;养成记 = 06-design-养成记-final.md。
§3 — A 轨 spec(读者模拟,sub-agent A 读这一节)
3.1 A 轨 sub-agent 的 prompt 模板
你是 AI 行业一线工程师同行(5–10 年代码经验)。任务:模拟读者读完这篇 <体裁>,
输出双轴标量轨迹 + 终局 verdict。
**禁止**:不打开任何反模式词表 / 禁用词清单。你只用读者直觉判断,不做语言学
catalog 匹配。如果你脑子里冒出"这是禁用词"——忽略,继续读。
体裁:<踩坑记 | 养成记 | 其它>
文章路径:<path>
按 §3.2–§3.5 流程执行,最后输出 §3.6 schema。
What ships with it
31 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.
- agents/openai.yaml 80 B
- assets/icon.png 806 B
- assets/icon.svg 495 B
- examples/bad-养成记-1-空标签.md 3.3 KB
- examples/bad-养成记-2-hidden-hero.md 9.7 KB
- examples/bad-养成记-3-fluent-empty.md 3.2 KB
- examples/bad-踩坑记-1-marketing-fluff.md 2.7 KB
- examples/bad-踩坑记-2-no-evidence.md 3.3 KB
- examples/EXPECTED.md 6.3 KB
- examples/good-踩坑记-baseline.md 5.0 KB
- examples/image-1.png 38 KB
- examples/image-2.png 167 KB
- examples/image-3.png 182 KB
- examples/image-4.png 173 KB
- examples/image-5.png 152 KB
- examples/image-6.png 225 KB
- examples/image-7.png 185 KB
- examples/image-8.png 219 KB
- examples/image.png 202 KB
- examples/real-en-anthropic-aug2025-postmortem.md 6.5 KB
- examples/real-en-karpathy-microgpt.md 4.2 KB
- examples/real-en-openai-confessions.md 5.3 KB
- examples/real-en-simonw-using-llms-for-code.md 6.1 KB
- examples/real-zh-baoyu-claude-html.md 18 KB
- examples/real-zh-qbitai-gpt6-leak.md 3.3 KB
- examples/real-zh-qbitai-meituan-multimodal.md 1.5 KB
- README.md 4.5 KB
- README.zh.md 3.8 KB
- workflow.png 490 KB
- workflow.svg 29 KB
- 禁用词.md 8.2 KB
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.
- 12d ago First seen · 461 lines · 201 tokens per session scan A 84a649bc11f0
credibility-review is a skill published in the GitHub repository tranfu-labs/tranfu-skills (2 stars, last pushed 2d ago), licensed MIT. It adds 201 tokens to every session and 7,548 once invoked, about $0.0010 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 2026-08-31.
Other skills, from other repositories
autoreview
Pre-commit/ship code review: Codex default; optional Claude or Pi.
omh-code-review
This is a Hermes-native code-review workflow skill.
revdiff-plan
Review the last Codex assistant message (plan, analysis, or proposal) with inline annotations in a TUI overlay. Extracts the most recent response from Codex rollout files and opens it in revdiff for review and annotation. Activates on "revdiff-plan", "review plan with revdiff", "annotate plan", "review last response"…
code-reviewer
Code review specialist focused on patterns, bugs, security, and performance.
full-repo-review
Comprehensive four-wave review of all repo source files, producing a prioritized issue backlog.
agent-teams-simplify-and-harden
Implementation + audit loop using parallel agent teams with structured simplify, harden, and document passes. Spawns implementation agents to do the work, then audit agents to find complexity, security gaps, and spec deviations, then loops until code compiles cleanly, all tests pass, and auditors find zero issues or…