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/kangarooking/loop-engineering-skill/comprehension-gapnpx skills add kangarooking/loop-engineering-skill --skill comprehension-gapgit clone --depth 1 https://github.com/kangarooking/loop-engineering-skillWhat 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 | $0.00127 | $0.01518 |
| Opus 5 | $0.00063 | $0.00759 |
| Sonnet 5 | $0.00025 | $0.00304 |
| Haiku 4.5 | $0.00013 | $0.00152 |
Grade A, and why
comprehension-gap 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 2d 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 — 107 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Source Metadata
Original cangjie-skill frontmatter from the distillation run:
name: comprehension-gap
description: |
识别和管理 Loop 系统的认知风险: 自动化程度越高,你真正理解的部分越少。
当用户发现"仓库里有很多代码但我看不懂"、"loop 产出很多但不确定对不对"、或需要向团队警示 AI 使用风险时使用。
不适用于: 纯技术调试、或产出完全可客观验证的场景。
关键 trigger: "AI 生成的代码我看不懂"、"产出太多理解不过来"、"AI 使用风险"。
source_book: "Loop Engineering 视频合集"
source_chapter: 视频3 (小木头) / 视频4 (Idoos Money)
tags: [risk, cognition, understanding, safety, human-in-the-loop]
related_skills: [loop-worthiness-test, maker-checker, three-stage-evolution]
认知差距警告 — Loop 的隐性风险
R — Reading (原文)
"理解在loop越快的交付你没有亲手写的代码,亲手看的代码,仓库里的东西和你脑子里真正搞懂东西差距就越大。另一个更加扎心的事实是最危险的姿态是舒舒服服的接受loop吐出来的一切,这是非常危险的。在AI生成的内容中,越多我们不了解的内容,对于我们来讲风险就越大。" — 小木头 (视频3)
"无人盯着的loop,也是无人盯着的在犯错。所以验证这个事情,永远还是在你自己手上。" — 小木头 (视频3)
I — Interpretation (自述)
核心洞察: 自动化程度与理解深度负相关。
Loop 交付速度越快 → 你没亲手写/看的代码越多 → 仓库里有的东西和你真正搞懂的东西之间的差距越大 → 风险越高。
最危险的态度: 舒舒服服地接受 Loop 输出的一切,不加审视。
关键原则: 验证永远在你自己手上。即使 loop 有自我验证环节,人仍需定期审查。
A1 — Past Application (书中案例)
案例1: Ralph Loop 失败 (视频3)
- 一个出了名的循环: 锲而不舍,永不放弃
- 失败模式: 无人盯着的 loop 持续犯错,把小修复变成灾难
- 教训: 没有人类监督的 loop 会制造"理解鸿沟"
案例2: Boris 的代码库 (视频2)
- Loop 产出的代码占仓库很大比例
- Boris 强调"代码库必须对 agent 可读" — 但可读 ≠ 人理解
- 隐含风险: agent 能改代码 ≠ 团队能理解为什么这么改
A2 — Future Trigger (未来触发)
- 产出理解不过来时: "仓库里有很多代码但我看不懂"
- Loop 产出质量不确定时: "它说做完了,但我不确定对不对"
- 团队 AI 使用风险警示时: 向团队传达"自动化≠理解"
- 设计 Loop 监督机制时: 在哪些环节加入人类审查?
语言信号: "AI 生成的代码我看不懂"、"产出太多理解不过来"、"AI 使用风险"、"无人盯着的 loop"
与相邻 skill 的区别:
loop-worthiness-test: 判断要不要做 loop (本 skill 是运行后的风险管理)maker-checker: 用 AI 审查 AI 的质量 (本 skill 是人类理解层面的风险)three-stage-evolution: 阶段定位 (本 skill 是 Stage 3 的特定风险)
E — Execution (可执行步骤)
Step 1: 评估认知差距
定期问自己:
- 这个 loop 产出的东西,我能解释给新人听吗?
- 如果 loop 出错,我能快速定位问题吗?
- 仓库里有多少代码/内容是我没亲手看过的?
Step 2: 设置人类审查节点
- 定期审查: 每周/每月审查 loop 产出样本
- 关键节点审查: 在 loop 重大变更、新任务类型、或异常指标时触发人工审查
- 理解度检查: 随机抽取 loop 产出,尝试用自己的话解释其逻辑
Step 3: 缓解策略
- 文档化: 要求 loop 在产出时附带"设计说明"
- 渐进式自动化: 从
loop-build-path的 Step 1 开始,确保每步都理解后再升级 - 团队 review: 定期让团队成员互相审查 loop 产出
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.
- 2d ago First seen · 107 lines · 127 tokens per session scan A b56da8e01b10
comprehension-gap is a skill published in the GitHub repository kangarooking/loop-engineering-skill (23 stars, last pushed 2mo ago), licensed MIT. It adds 127 tokens to every session and 1,518 once invoked, about $0.0006 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
mcporter
List, auth, and call MCP servers/tools from the terminal.
📝 任务完成后归档
重要提醒: 每次完成复杂调试或开发任务后，主动执行此流程！ 将学到的经验归档为 skill，供以后参考。不要等用户提醒。.
oracle
Best practices for using the oracle CLI (prompt + file bundling, engines, sessions, and file attachment patterns).
agent-mode
Unified tool for managing agent LLM modes (add, remove, update, list, switch).
agento11y-prod-setup
Sets up production evaluation and guardrails for a DEPLOYED AI agent in Grafana Agent Observability, grounded in the agent's own code and its real ingested traffic. The judgment layer on top of the agento11y skill: it reads the agent's source (system prompt, tools, entrypoint) AND samples its live traffic via gcx…
mcp-scripting
Write mcpScript JavaScript for discovering, inspecting, and calling MCP tools.