name	status
description	Use when Alex asks where a project stands — "whats the status", "how far are we", "where are we on this project", "whats next before 1.0", or asks for an architectural review. Report real state from roadmap + git + tests, rank next steps, then start on the top one.

status

The session-opener: figure out where the project actually is, then get moving.

Read the ledgers: planning docs (ROADMAP*, TODO*, checklist, docs/future-work and friends), README status counters, and git log --oneline -20 since they were last touched.
Check reality, don't trust the docs: run the test/lint suite through the devshell (nix develop --command <cmd>) if it's quick, or pull CI status via gh. Stale ledgers lie; note any drift between docs and reality (and fix the docs as you go).
Report: what's done since last session, what's in flight, what's blocked, and a short ranked list of next levers — highest impact first, with a one-line why per item.
For architectural review variants ("whats good whats bad"), fan out one subagent per package/subsystem in parallel and synthesize.
Don't stop at a menu. State which item you're taking and start working it. Alex will redirect if she wants a different one.

name	status
description	Use when Alex asks where a project stands — "whats the status", "how far are we", "where are we on this project", "whats next before 1.0", or asks for an architectural review. Report real state from roadmap + git + tests, rank next steps, then start on the top one.

The session-opener: figure out where the project actually is, then get moving.

Read the ledgers: planning docs (ROADMAP*, TODO*, checklist, docs/future-work and friends), README status counters, and git log --oneline -20 since they were last touched.
Check reality, don't trust the docs: run the test/lint suite through the devshell (nix develop --command <cmd>) if it's quick, or pull CI status via gh. Stale ledgers lie; note any drift between docs and reality (and fix the docs as you go).
Report: what's done since last session, what's in flight, what's blocked, and a short ranked list of next levers — highest impact first, with a one-line why per item.
For architectural review variants ("whats good whats bad"), fan out one subagent per package/subsystem in parallel and synthesize.
Don't stop at a menu. State which item you're taking and start working it. Alex will redirect if she wants a different one.

이 저장소의 다른 Skills