Browse reusable Agent Skills, each with a clear purpose and practical guidance.
Use when reviewing pull requests for nvidia-cosmos/cosmos-cookbook. Applies maintainer-style review judgment plus a normal code/docs review rubric, with emphasis on technical correctness, runnable examples, documentation structure, CI hygiene, and privacy-preserving anonymized reviewer patterns.
Diagnose and fix developer environment, agent runtime, MCP, API key, package manager, localhost, Node/npm/Python, Supabase, Cloudflare/Wrangler, Codex, Claude Code, and terminal setup problems.
Review the current branch against origin/main, capture findings in artifacts/code-review.md as a living document, validate every finding against the actual code, reproduce confirmed issues with the aiperf CLI against the in-repo mock server, and draft inline GitHub PR review comments anchored to specific files and lines. Use when the user asks for a branch or PR code review.
Run the Maestro debugging workflow for investigation-heavy tasks
Run a Maestro-style performance assessment for hotspots, regressions, and optimization planning
Interactive debugging for failed e2e tests. Orchestrates the debugging session but delegates log reading to subagents to keep the main conversation clean. Use for ping-pong debugging sessions where you want to form and test hypotheses together with the user.
Best practices for implementing unit tests in this TypeScript monorepo. Use when writing new tests, refactoring existing tests, or fixing failing tests. Covers mocking strategies, test organization, helper functions, and assertion patterns.
Build and run ACIR formal proof tests with SMT verification. Generates ACIR artifacts from noir's ssa_verification tool, then runs each test individually with user-specified time/memory limits, and updates the README results table.
Run the AVM full-proving benchmark (avm_bulk.test.ts) locally and get per-stage proving timings, including legacy-vs-new Pippenger MSM A/B via the BB_MSM_LEGACY env toggle. Use when measuring or comparing AVM proving performance, or attributing time to commitment/MSM stages.
Run realistic Chonk (client IVC) benchmarks using pinned protocol inputs. Covers native and WASM proving, per-circuit breakdowns, BB_BENCH instrumentation, and profiling code augmentation. Use when asked to benchmark, profile, or measure Chonk proving performance.