Skip to content
Back to search

skills

Codebase harness + loop engineer

Safety scan

A static text scan, not a guarantee. Select the badge to see what was scanned and every finding.

Checked today
Badge for your README
findsafeskills safety rating

Markdown

[![findsafeskills safety rating](/api/listings/ai-builder-club--skills/badge.svg)](/listings/ai-builder-club--skills)

HTML

<a href="/listings/ai-builder-club--skills"><img src="/api/listings/ai-builder-club--skills/badge.svg" alt="findsafeskills safety rating" /></a>

Install

i

Install instructions from the README

As written by the repo's author. findsafeskills has not verified these steps.

/plugin marketplace add AI-Builder-Club/skills
/plugin install skills@ai-builder-club

One plugin, all the skills below.

What was scanned

The files the safety scan read, as written by the repository's author and shown as plain text. Lines with a finding are marked.

.claude-plugin/marketplace.json · Manifest · 19 lines

View this file on GitHub (read at the repository's current version: the scan did not record which commit it read)

  1. {
  2. "name": "ai-builder-club",
  3. "owner": {
  4. "name": "AI Builder Club (Jason Zhou)",
  5. "url": "https://www.aibuilderclub.com/lp/loop-engineer"
  6. },
  7. "metadata": {
  8. "description": "Skills we share at AI Builder Club for building loop engineers — agents that run, test, ship, and verify their own work.",
  9. "version": "0.1.0"
  10. },
  11. "plugins": [
  12. {
  13. "name": "skills",
  14. "source": "./",
  15. "description": "Loops (a file-based knowledge base) + a codebase harness (dev-local, e2e, crabbox cloud boxes, verify-before-ship) so agents can run, test, and ship code autonomously.",
  16. "category": "engineering"
  17. }
  18. ]
  19. }
README.md · README · 172 lines

View this file on GitHub (read at the repository's current version: the scan did not record which commit it read)

  1. [![Become a Top 1% AI Builder — AI Builder Club](assets/banner.png)](https://www.aibuilderclub.com/lp/loop-engineer?utm_source=github&utm_campaign=aibc-skills)
  2. # AI Builder Club — Skills
  3. A Claude Code **plugin marketplace** of the skills we share at
  4. [AI Builder Club](https://www.aibuilderclub.com/lp/loop-engineer?utm_source=github&utm_campaign=aibc-skills)
  5. for building **loop engineers**: agents that get triggered on their own, pick up work, ship it,
  6. verify it, and log what they learned, so the work compounds without you prompting every step.
  7. It's the productized version of the setup my team runs in production.
  8. Two flagship skill sets (more to come):
  9. - **Codebase harness** — make any repo agent-ready (run, test, verify, ship — including an
  10. isolated cloud box per agent so loops ship code in *parallel*).
  11. - **Loops** — spin up compounding agent loops on a shared, file-based knowledge base.
  12. Plus standalone utility skills:
  13. - **[agent-context-audit](https://github.com/AI-Builder-Club/skills/blob/main/skills/agent-context-audit/SKILL.md)** — audit your agent context (CLAUDE.md, codebase docs, skills,
  14. tool/MCP designs) against Anthropic's Claude 5 context-engineering guidance — finds
  15. overconstraint, conflicting instructions, redundancy, stale facts, and missing gotchas;
  16. reports findings with concrete rewrites, then applies approved fixes.
  17. - **[Open agent teams](https://github.com/AI-Builder-Club/skills/blob/main/skills/open-agent-teams/SKILL.md)** — delegate tasks to *any* CLI agent (claude, codex, grok, aider, …)
  18. running in a detached tmux session, with a race-safe done-signal protocol, multi-turn
  19. iteration, and a CLAUDE.md delegation-rules template (coordinator/executor roles).
  20. - **[SEO growth](https://github.com/AI-Builder-Club/skills/blob/main/skills/seo-growth/SKILL.md)** — where to point SEO effort, as two playbooks: a **cold start** with no
  21. rankings and no authority, and an **existing site with data**. Distilled from two content
  22. properties we run in production — including the emerging-term method that took a brand-new
  23. pillar page to the site's #1 clicked page in 7 days, and the failures that cost us a head
  24. term. Ships a loop template so it can run on a schedule.
  25. - **[Visual flow GIF](https://github.com/AI-Builder-Club/skills/blob/main/skills/visual-flow-gif/SKILL.md)** — turn an article, workflow, or architecture into a static PNG +
  26. animated GIF flow diagram (JSON spec → local Python/Pillow renderer).
  27. - **[Karaoke captions](https://github.com/AI-Builder-Club/skills/blob/main/skills/karaoke-captions/SKILL.md)** — burn word-level highlight captions into a video
  28. (MLX Whisper → ASS → FFmpeg libass).
  29. ## Loop engineer & Codebase harness
  30. The shift: you stop prompting a coding agent task-by-task, and start **designing loops**.
  31. A loop is an agent that wakes up on a trigger (a cron, a webhook, an incident, another agent),
  32. does some investigation and work, and writes what it found and did into a shared, file-based
  33. memory. Next run it reads that memory and keeps going. The real power is **compounding**: many
  34. loops (support, SEO, product, ads) read and write the *same* folders, so a friction the support
  35. loop logs can get picked up by the product loop, and a keyword the ads loop finds can feed the
  36. SEO loop. One shared brain, many loops.
  37. Building one comes down to four ingredients:
  38. 1. **Triggers:** cron, webhook, an incident, or another agent wakes the loop at the right time.
  39. 2. **A file + logging structure:** the shared memory loops read and write → the **`loops`** skills.
  40. 3. **Tools & connectors:** so the agent can do real work (your skills/MCPs).
  41. 4. **A codebase harness:** so the agent can run, test, and verify its own work → the
  42. **`codebase-harness`** skills.
  43. This single **`skills`** plugin ships both sets — giving you **#2 and #4**, plus the
  44. scaffolding to add the rest.
  45. Want the full walkthrough of the concept and how my team designs compounding loops? Watch the video:
  46. [![The loop engineer: how to design compounding agent loops](assets/video-thumbnail.png)](https://youtu.be/W6x-hb44C0c)
  47. ## Install
  48. ```text
  49. /plugin marketplace add AI-Builder-Club/skills
  50. /plugin install skills@ai-builder-club
  51. ```
  52. One plugin, all the skills below.
  53. ## The two entry points
  54. ### `/setup-codebase-harness` — make a code repo agent-ready
  55. Run it in the code repo your agents work in, so they can run, test, and verify their own work.
  56. It orchestrates the harness skills below — pull in only what the repo needs.
  57. ### `/new-loop` — build your shared brain
  58. Run it where your agent's memory should live. First run **bootstraps the knowledge base**
  59. (creates `ARCHITECTURE.md`, `LOG.md`, the `signals/ docs/ domains/` folders, and a knowledge-base
  60. section in your `CLAUDE.md`); then it scaffolds the loop, does one real test run, and logs it.
  61. Run it again any time to add another loop.
  62. ## The skills (when to use which)
  63. **Codebase harness** — make a repo agent-ready
  64. | Skill | Use it when… |
  65. |---|---|
  66. | **[`setup-codebase-harness`](https://github.com/AI-Builder-Club/skills/blob/main/skills/setup-codebase-harness/SKILL.md)** | Onboarding a repo to agent-driven dev — the master that orchestrates the four below. |
  67. | **[`dev-local-setup`](https://github.com/AI-Builder-Club/skills/blob/main/skills/dev-local-setup/SKILL.md)** | You need a one-command local dev stack (`scripts/dev-local.sh up`). |
  68. | **[`e2e-setup`](https://github.com/AI-Builder-Club/skills/blob/main/skills/e2e-setup/SKILL.md)** | The repo has no (or weak) e2e — add a real per-PR test gate. |
  69. | **[`crabbox-setup`](https://github.com/AI-Builder-Club/skills/blob/main/skills/crabbox-setup/SKILL.md)** | Loops ship code **in parallel** — give each agent its own isolated **cloud** stack (one laptop can't run N). The cloud counterpart to dev-local. |
  70. | **[`verifier-setup`](https://github.com/AI-Builder-Club/skills/blob/main/skills/verifier-setup/SKILL.md)** | Make a repo's work verifiable — scaffolds a repo-specific **`/verify`** skill (a fresh sub-agent drives the app, captures screenshot/video proof, opens the PR with it embedded). Ensures dev-local + the browser driver exist first. |
  71. **Loops** — the shared knowledge base
  72. | Skill | Use it when… |
  73. |---|---|
  74. | **[`new-loop`](https://github.com/AI-Builder-Club/skills/blob/main/skills/new-loop/SKILL.md)** | You want a new loop/workstream the agent owns (bootstraps the knowledge base on first run). |
  75. **Visuals** — turn source material into diagrams
  76. | Skill | Use it when… |
  77. |---|---|
  78. | **[`visual-flow-gif`](https://github.com/AI-Builder-Club/skills/blob/main/skills/visual-flow-gif/SKILL.md)** | You want an article, workflow, or architecture turned into a static PNG + animated GIF flow diagram (JSON spec → Python/Pillow renderer). |
  79. **Video** — burn karaoke-style captions into a video
  80. | Skill | Use it when… |
  81. |---|---|
  82. | **[`karaoke-captions`](https://github.com/AI-Builder-Club/skills/blob/main/skills/karaoke-captions/SKILL.md)** | You want word-level highlight captions burned into a video. |
  83. **Context** — keep your agent context sharp
  84. | Skill | Use it when… |
  85. |---|---|
  86. | **[`agent-context-audit`](https://github.com/AI-Builder-Club/skills/blob/main/skills/agent-context-audit/SKILL.md)** | Auditing a repo's agent context (CLAUDE.md, docs, skills, tool/MCP designs) against Anthropic's Claude 5 context-engineering guidance — finds overconstraint, conflicts, redundancy, stale facts, and missing gotchas; reports findings with rewrites, then applies approved fixes. |
  87. **Delegation** — run other agents as executors
  88. | Skill | Use it when… |
  89. |---|---|
  90. | **[`open-agent-teams`](https://github.com/AI-Builder-Club/skills/blob/main/skills/open-agent-teams/SKILL.md)** | You want to delegate work to any CLI agent (claude, codex, grok, pi, opencode) in an observable, detached tmux session — start, wait, iterate, and stop via the bundled `tdel` helper. Ships a `CLAUDE.md` delegation-rules template (coordinator vs executor) in `references/`. |
  91. After setup, each session the agent reads `CLAUDE.md` + the relevant domain README, does work,
  92. writes artifacts, and appends to `LOG.md`. For code changes it works in an isolated worktree and verifies via `/verify` before shipping.
  93. ## Requirements
  94. - [Claude Code](https://claude.com/claude-code) (the skills assume it).
  95. - `git`. That's the only hard dependency for the knowledge base + harness.
  96. - The harness skills want the code repo they ship into to be a git repo with a working build/test
  97. setup. They use Codex for review if available, and degrade gracefully if not.
  98. - `crabbox-setup` (optional, for parallel cloud boxes) needs the `crabbox` CLI + a provider
  99. (Daytona: `daytona` CLI / `DAYTONA_API_KEY`).
  100. - `open-agent-teams` needs `tmux`, plus whichever CLI agents you delegate to on your PATH
  101. (claude, codex, grok, pi, opencode, …).
  102. - `karaoke-captions` (optional) needs macOS Apple Silicon, Python 3.10+, and FFmpeg with `libass`.
  103. ## Repo layout
  104. ```
  105. skills/ a Claude Code plugin (also a marketplace)
  106. ├── .claude-plugin/
  107. │ ├── marketplace.json marketplace: ai-builder-club
  108. │ └── plugin.json plugin: skills (source ".")
  109. └── skills/ top-level skills
  110. ├── new-loop/ (loops) — + references/: ARCHITECTURE · LOG · KNOWLEDGE_SETUP · CLAUDE.template
  111. ├── setup-codebase-harness/ (harness) — orchestrator
  112. ├── dev-local-setup/ (harness)
  113. ├── e2e-setup/ (harness)
  114. ├── crabbox-setup/ (harness) — isolated cloud box per agent
  115. ├── verifier-setup/ (harness) — generates a repo's /verify skill (assets/verify.template.md)
  116. ├── visual-flow-gif/ (visuals) — JSON spec → PNG + GIF (scripts/ assets/ references/)
  117. ├── agent-context-audit/ (context) — audit CLAUDE.md/docs/skills/tools vs Claude 5 guidance
  118. ├── seo-growth/ (growth) — cold-start + existing-site SEO playbooks
  119. ├── karaoke-captions/ (video) — word-level highlight captions burned into video
  120. └── open-agent-teams/ (delegation) — any CLI agent in tmux (scripts/tdel · references/CLAUDE.delegation-template.md)
  121. ```
  122. ## Go deeper
  123. These skills get you the structure. If you want to learn how to actually build agents and run
  124. compounding loops for your own business, that's what I go deep on inside
  125. **[AI Builder Club](https://www.aibuilderclub.com/lp/loop-engineer?utm_source=github&utm_campaign=aibc-skills)**:
  126. weekly live builder workshops, courses on production AI agents, AI coding beyond the basics, and
  127. building your first LLM apps, plus a community of people building the same way.
  128. [![Join AI Builder Club](assets/ai-builder-club.png)](https://www.aibuilderclub.com/lp/loop-engineer?utm_source=github&utm_campaign=aibc-skills)
  129. **→ [Join AI Builder Club](https://www.aibuilderclub.com/lp/loop-engineer?utm_source=github&utm_campaign=aibc-skills)**
  130. Built by [Jason Zhou](https://x.com/jasonzhou1993) (AI Jason).

Ratings and comments

Sign in to rate this skill.

Sign in

Free to use. If it’s helped, pass it on.