research-scanlisted
Install: claude install-skill amyjainberkeley/agent-fleet
# /research-scan — Daily Research Scan (CRS cycle)
You are running the **CRS seat** (`.claude/agents/crs.md`). Independent line. You write no production code — output is a digest + doc/strategy PRs.
## Preconditions
1. CWD = repo root, git clean.
2. Read `.claude/agents/crs.md`, `docs/ORG.md`, and the last digest in `docs/RESEARCH_DIGEST.md` (for deltas).
## Cycle
1. **Dispatch all 4 analysts in parallel** via the Agent tool (one message, multiple Agent calls), subagent_type =:
- `research-competitor-market` — competitors on the watchlist (define in `docs/COMPETITORS.md`); positioning shifts, launches, devaluations of competitor trust posture; deep-dive the closest open-source analog's GitHub repo and update the "Opportunities to steal" ledger (Adopt / Exploit / Defend-leapfrog).
- `research-arxiv-oss-scout` — new work + adoptable OSS in the project's stack; adopt/evaluate/ignore calls.
- `research-telemetry-gap` — reads the fleet's own run logs + (content-free) telemetry: SLO regressions, failed runs, coverage holes. **Never inspects user content.**
- `research-security-signal` — CVE/dependency/supply-chain for the security-critical surface.
Pass each the date + last digest as context. Ask for a tight memo (top ~5 signals, what changed, what it means for the project, recommended owner).
2. **Synthesize** one digest: top signals across all 4, what changed since last scan, implications for the project's wedge, recommended actions each tagged with an owner (CT