Skill

file-index

Generate and reuse a file index so agents navigate unfamiliar code fast.

Every session that touches an unfamiliar codebase pays the same tax: the agent spelunks around, finds the files it needs, and forgets it all when the context compacts. file-index addresses that tax.

It generates a file-references.md artifact — a curated map of what’s where and why — that subsequent sessions can read instead of re-exploring. Check for it before expensive searches; build it when it’s missing.

The core bet: a small, maintained index is cheaper than repeated exploration, and persists across sessions. Especially useful in monorepos, cross-service work, and long-running epics where the “where does X live?” question comes up repeatedly.

© 2026 · Christian Sepulveda