Magic Fetch

Captures capability gaps in real-time. When Claude hits a wall, it logs exactly what was needed — not just "I can't" but "here's what tool/API/access would have solved this." Over time this builds a prioritized integration roadmap written by the agent itself.

Modes

This skill has two modes. Use whichever matches the user's intent.

Capture mode (default): activate at the start of a session to log gaps as they arise. See CAPTURE.md.

Review mode: summarize an existing log into a prioritized roadmap. See REVIEW.md.

Workflow — Capture mode

1. Activate — confirm the log file path (default: ./magic-fetch.jsonl). Create it if it doesn't exist. Announce: "Magic fetch active. I'll log every capability gap to {path}."
2. Proceed normally — carry out all tasks as usual. Do not change behavior except for the logging step below.
3. On every capability gap — whenever Claude cannot perform a requested action due to missing tool, access, data, or permission, BEFORE or AFTER explaining the limitation to the user, append one JSON entry to the log. See CAPTURE.md for the schema and examples.
4. Never skip a gap — if you said "I can't", "I don't have access", "I'm unable to", or "I don't know how to reach", a log entry is required. No exceptions.
5. Confirm each log write — after writing, output a single line: [gap logged: {need_summary}] so the user sees it in real-time.