AI Builder Club — Skills

A Claude Code plugin marketplace of the skills we share at AI Builder Club for building loop engineers: agents that get triggered on their own, pick up work, ship it, verify it, and log what they learned, so the work compounds without you prompting every step. It's the productized version of the setup my team runs in production.

Two flagship skill sets (more to come):

• Codebase harness — make any repo agent-ready (run, test, verify, ship — including an isolated cloud box per agent so loops ship code in parallel).
• Loops — spin up compounding agent loops on a shared, file-based knowledge base.

Plus standalone utility skills:

• agent-context-audit — audit your agent context (CLAUDE.md, codebase docs, skills, tool/MCP designs) against Anthropic's Claude 5 context-engineering guidance — finds overconstraint, conflicting instructions, redundancy, stale facts, and missing gotchas; reports findings with concrete rewrites, then applies approved fixes.
• Open agent teams — delegate tasks to any CLI agent (claude, codex, grok, aider, …) running in a detached tmux session, with a race-safe done-signal protocol, multi-turn iteration, and a CLAUDE.md delegation-rules template (coordinator/executor roles).
• SEO growth — where to point SEO effort, as two playbooks: a cold start with no rankings and no authority, and an existing site with data. Distilled from two content properties we run in production — including the emerging-term method that took a brand-new pillar page to the site's #1 clicked page in 7 days, and the failures that cost us a head term. Ships a loop template so it can run on a schedule.
• Visual flow GIF — turn an article, workflow, or architecture into a static PNG + animated GIF flow diagram (JSON spec → local Python/Pillow renderer).
• Karaoke captions — burn word-level highlight captions into a video (MLX Whisper → ASS → FFmpeg libass).

Loop engineer & Codebase harness

The shift: you stop prompting a coding agent task-by-task, and start designing loops.

A loop is an agent that wakes up on a trigger (a cron, a webhook, an incident, another agent), does some investigation and work, and writes what it found and did into a shared, file-based memory. Next run it reads that memory and keeps going. The real power is compounding: many loops (support, SEO, product, ads) read and write the same folders, so a friction the support loop logs can get picked up by the product loop, and a keyword the ads loop finds can feed the SEO loop. One shared brain, many loops.

Building one comes down to four ingredients: