dev

Core skills library: TDD, debugging, brainstorming, plan-writing, code-review patterns.

Git worktree management guidance and activity-tracking hooks for parallel AI agent workflows.

Pause execution and inspect runtime state via DAP. Attach to live processes by PID, no restart or code changes required.

Evaluate Claude Code skills against best practices. Tracks skill usage globally via hooks.

AI-powered codebase analysis that produces interactive knowledge graphs, guided tours, and plain-English architecture explanations.

Full ElevenLabs coverage for building with the API: the 9 official ElevenLabs agent skills (vendored, re-syncable) plus gap skills that pull live docs indexes for voices, dubbing, dialogue, studio, admin, SDKs, image-video, Reception AI, and a docs-lookup catch-all.

Agent skills for Qdrant vector search: scaling, performance, search quality, monitoring, deployment, SDK usage.

Spawn a peer back-and-forth between Claude (max thinking) and Codex (high reasoning) on hard architecture decisions.

Routes all delegable work (exploration, bulk reads, tests, mechanical edits, first-pass reviews, sweeps) from a Fable session to the Codex CLI via codex exec, spending Codex quota instead of Anthropic tokens.

Pull-request review doctrine (triage, lenses, severity, stop rules) plus agent-specific tactics (context budget, tool escalation, review fan-out, responsible non-review).