Fans out one background agent per open issue/PR to produce evidence-backed, permalink-cited triage reports in /tmp — with zero writes to GitHub.
Case-by-case QA of opencode itself — CLI, HTTP server, plugin hooks/events over SSE, tmux TUI smoke, and read-only session DB investigation.
An install playbook that adds the pinned Firecrawl AnyDoc CLI to NanoClaw agent containers so Office documents convert to Markdown locally.
QA the omo Codex plugin in a fully isolated CODEX_HOME with a mock model, proving hooks actually fire without ever touching the user's real ~/.codex.
Splits a task into atomic PRs, builds each in its own git worktree, and loops through CI and bot review gates until merged.
Runs 3 vulnerability hunters and 2 PoC engineers in parallel so only findings with a proven, reproducible attack path survive.
A step-by-step runbook for wiring DeltaChat — email-based, end-to-end encrypted messaging — into a NanoClaw bot.
Runs a three-layer, parallel multi-agent review of all unpublished changes before an npm publish and returns a SAFE/CAUTION/RISKY/BLOCK verdict plus a version-bump recommendation.
Official skill that walks you through researching, implementing, testing, and evaluating high-quality MCP (Model Context Protocol) servers in TypeScript or Python.
Designs and runs failure-injection experiments for distributed systems, with blast-radius controls, rollback scripts, and game day playbooks.
Reads your local screenpipe timeline, clusters repeated workflows, and drafts new skills or composition recipes for the patterns not yet covered.
Guides Claude to infer transcription factor–target gene networks from expression data using arboreto's GRNBoost2 and GENIE3.
Look up precomputed AlphaGenome Atlas AVI scores and call the AlphaGenome model on demand to rank and mechanistically interpret non-coding regulatory SNVs.
Convenes each git worktree and PR as its own agent in a shared markdown room, then drives rounds until one consolidation plan emerges.
Maps a codebase into feature-grouped Mermaid flowcharts, hunts duplicated concerns with file:line evidence, and proposes the simplest unified architecture before a refactor.
Typing 'ultrawork' or 'ulw' flips the agent into a strict failing-first, evidence-driven engineering mode with mandatory cleanup receipts.
Triggered by the keyword 'ultrawork' or 'ulw', it forces a failing-proof-first, real-surface-QA, cleanup-receipt workflow before anything can be called done.
Runs a named team of cooperating Codex workers under the main session as leader, with durable on-disk team state.
Debugs a LazyCodex or upstream Codex CLI defect in a throwaway workspace and ships it as a fork PR or a patch-embedded verified-fix issue.