improve-api-from-learnings

name	improve-api-from-learnings
description	Use when improving Nevermind's extension API or builder workflow from accumulated learning traces. Export learnings first, then analyze the raw artifacts to propose API, prompt, or builder changes.

Improve API From Learnings

Use this skill when the goal is to improve Nevermind itself from past extension-building friction, not when building a user extension.

Workflow:

1. Run mise exec pnpm -- pnpm learnings:export.
2. Read the newest export bundle under .tmp/learnings-export/.
3. Treat traces.json as the main evidence stream.
4. Treat learnings.md and learnings.json as the current user-learning output, not the source of truth for product changes.
5. Identify repeated friction in how extensions are built: missing API primitives, confusing names, prompt failures, weak tool descriptions, repeated retries, or host gaps.
6. Prefer improving src/resources/nevermind-extension-api.d.ts, src/electron/ai.ts, the builder skill, or host behavior at the source instead of adding workaround instructions.
7. Make small reviewable changes and verify with mise exec pnpm -- pnpm test.

Rules:

• Do not use this skill for user-extension authoring; use the normal extension-builder flow for that.
• Do not treat one-off chat details as product learnings; look for repeated or clearly generalizable friction.
• Do not bloat prompts with many new rules; prefer fixing API shape, tool descriptions, or host primitives.
• When prompt changes are needed, tighten or replace existing guidance instead of appending parallel instructions.
• When API changes are needed, update src/resources/nevermind-extension-api.d.ts as the canonical contract.
• Keep developer learnings grounded in raw exported traces, especially tool calls, validation cycles, and runtime failures.