research-statuslisted

Use when the lead architect asks for current project progress, global research state, key bottlenecks, current architecture, active research, or a compact project context snapshot. Compiles canonical repository state into a Chinese Project Working Model with traceable evidence. Read-mostly synthesis — it does not plan research, start experiments, or modify state.
uukuguy/research-engineering · ★ 0 · AI & Automation · score 72

Install: claude install-skill uukuguy/research-engineering

# Research Status Compile the repository's current research state into a compact Project Working Model. Two audiences read the same report: 1. the lead architect, who needs the global picture without reconstructing it; 2. an AI agent, which can paste it back in as high-quality context compression. It is a **derived snapshot**. Canonical truth stays in `research/ACTIVE.json`, `CURRENT.md`, `ARCHITECT.md`, `BOUNDARIES.md`, `ENVIRONMENT.md`, `FINDINGS.md`, the evidence ledger, run manifests, and Git. The report never becomes a second source of truth. `AGENTS.md` carries the language policy, the authority split, and the invariants. This skill does not restate them. ## Language The visible report is **Chinese**, with technical terms, algorithm and library names, metrics, protocol names, Git refs and tags, and canonical IDs (`EV-*`, `EXP-*`, `H-*`, `FND-*`, `RB-*`, `ENV-*`) kept in their original English form. Keep the structure stable enough that the same report can be pasted into a new Codex or Claude Code session as usable context. Do not maintain a bilingual duplicate — one canonical English state, one Chinese report that references it by ID. ## Read-mostly By default this skill does **not**: - modify application code; - start experiments; - choose a new research direction; - change hypotheses or `FINDINGS.md`; - promote a candidate; - alter architect decisions; - reset, clean, or check out over Git state. If the state is inconsistent, report it, and then **name the