dustinkeeton/wafflestack

๐Ÿง‡ One batter, every repo โ€” reusable AI agent & skill definitions rendered into harness-native files (.claude/, .codex/, .agents/)

wafflestack๋ž€ ๋ฌด์—‡์ธ๊ฐ€์š”?

wafflestack is a Claude Code agent skill that ๐Ÿง‡ One batter, every repo โ€” reusable AI agent & skill definitions rendered into harness-native files (.claude/, .codex/, .agents/).

์ง€์› ๋Œ€์ƒโœ“Claude Codeโœ“Codex CLI~Cursor
npx skills add dustinkeeton/wafflestack

Installed? Explore more ๋ผ์ดํŒ… & ์—๋””ํŒ… skills: steipete/notion, langchain-ai/langchain, bytedance/podcast-generation ยท View all 6 โ†’

์ฆ๊ฒจ ์‚ฌ์šฉํ•˜๋Š” AI์—๊ฒŒ ๋ฌผ์–ด๋ณด๊ธฐ

์ด ์—์ด์ „ํŠธ ์Šคํ‚ฌ์ด ๋ฏธ๋ฆฌ ๋กœ๋“œ๋œ ์ƒˆ ์ฑ„ํŒ…์„ ์—ฝ๋‹ˆ๋‹ค.

๋ฌธ์„œ

Documentation Orchestration

When this skill is invoked, run the following pipeline:

Step 1: Architecture Audit

Spawn the general-purpose agent with this prompt:

Audit the current codebase for any changes since the docs were last updated. Read the existing machine docs (a single root AGENTS.md registry) and compare against the actual project. Produce a concise change report covering:

  • New or removed bundles, skills, agents, config keys, or CLI commands
  • Changed installer exports or template semantics (installer/lib/*.mjs)
  • bundle.yaml config declarations out of sync with placeholder usage
  • Rendered-output drift: node installer/cli.mjs doctor --allow-missing clean (committed render + lock in sync โ€” re-render and commit after bundle edits); .gitignore still covers the deliberately-untracked renders (label-hook workflow, worktrees, .waffle/ overview docs)

Output ONLY the change report as structured text. Do NOT modify any files.

Step 2: Agent Documentation

Take the change report and spawn the docs-agent agent with this prompt:

Here is the architecture change report for the wafflestack codebase:

{step-1 output}

Update the machine docs โ€” a single root AGENTS.md registry โ€” to reflect the current state of the project. Read the actual source files to verify details โ€” do not rely solely on the change report. Follow the docs-agent skill guidelines for format and structure.

Step 3: Human Documentation

After docs-agent completes, spawn the docs-human agent with this prompt:

The machine docs have just been updated. Update the human-facing documentation โ€” DECISIONS.md, STATUS.md, and ARCHITECTURE.md at the repo root โ€” to match.

Read the updated machine docs and the actual source to ensure accuracy. Follow the docs-human skill guidelines for format, structure, and its guardrails (owner-voiced docs are flagged, never rewritten).

Important

  • Each step must complete before the next begins โ€” the output of each informs the next.
  • Step 1 is read-only โ€” the audit agent only reports.
  • docs-agent and docs-human SHOULD modify their respective documentation files.

๊ด€๋ จ ์Šคํ‚ฌ