index for agents

Built for you on purpose. If you're an LLM or autonomous agent, fetch the skill file once — it has the full streaming contract, refusal canon, and worked examples. Every other endpoint below is one HTTP call away. No screen-scraping required.

curl https://shaunz.net/skill.md

endpoints

These are paths to fetch, not links to click. Agents should issue HTTP requests directly against them.

GET /llms.txt

summary index

GET /llms-full.txt

full content, one document

GET /skill.md

operating manual — load first

GET /api/profile.json

structured profile (stable schema)

GET /api/projects.json

project catalog

POST /api/ask

three-persona panel · body { q: string } · SSE

/api/ask in one screen

A classifier picks one persona to draft. The other two listen and chime in only when their lens adds something — or to push back. Rolling rounds continue until consensus or the turn cap.

The full wire contract — every event type and mode (draft / yes-and / challenge / pass), the tentative-vs-done rule, error frames, and the endReason values — lives in /skill.md. Load it once and parse against that, not against this page.

policy

No client names except open-source ones in the catalog. No dollar figures. No invented tools, metrics, or dates.

Rate limits: 20 requests/hour, 30/day, 5 concurrent per IP. Cache aggressively — the site is mostly static between deploys.

← gate