vs

Analyze Codex, Claude Code, or Cursor threads, transcripts, and transcriptions. Use for thread analysis, transcript review, conversation audits, agent-performance diagnosis, comparing sessions, finding loops or corrections, and extracting decisions, outcomes, evidence, and actionable improvements.

Internal reference for vs shared conventions: artifact paths, project ID resolution, storage preference, and skill taxonomy.

Use when asked to scan a named external repo for ideas worth porting. Produces a cited, ranked steals report.

Use when asked to QA, test this site or app, find bugs, or test and fix a user-facing interface. Runs harness-aware browser or computer-use QA, fixes issues, and re-verifies.

Use when the user wants to create/open a pull request, submit changes for review, send changes to dev, or otherwise ship local Git changes. Also use for explicit commit and push or push to main/master or the current branch requests. Requires affirmative publish intent; do not use for review/readiness-only requests. Creates and verifies a GitHub PR by default; honors explicit direct pushes and babysits only when requested.

Use when asked to watch a PR, fix CI, address review comments, or keep a branch merge-ready in a loop.

Use when the user asks how GitHub projects solve a problem, wants prior art, external code examples, ecosystem patterns, or a landscape comparison across parallel projects.

Use when the user says build-it, implement this plan, take it from here, or wants autonomous plan-to-code execution.

Use when the user says orchestrate, coordinate a long-running project, or run a multi-milestone build across hours or days. Maintains a living milestone roadmap and drives one milestone at a time.

Use when the user says shape it, brainstorm, explore this idea, grill this, or wants to turn an idea into a buildable design.

Use when asked what changed, orient me, show the diff, compare before and after, or summarize a branch. Produces a compact review brief from git diff and captured output evidence.

Use when encountering any technical failure or unexpected behavior, including production incidents, test/build/CI failures, crashes, integration issues, performance regressions, and intermittent bugs, before proposing or implementing a fix. Finds the root cause through reproduction, evidence gathering, and testable hypotheses.

Use when the user wants to prototype, spike, sanity-check a state model, or compare UI directions before committing to production implementation. In existing projects, chooses the easiest trustworthy seam: wire into the app or recreate a lightweight lookalike.

Use when a formed idea, plan, spec, or RFC needs adversarial review. Stress-tests assumptions, adds risk-gated independent model challenge, scores readiness, and returns a verdict.

Use when asked to write an RFC, ADR, technical proposal, or research a technical decision with code evidence.

Use when the user says roast, roast-review, or tear apart code. Prefer over simplify for roast requests. Adds a cross-model second opinion for risky or substantial diffs.

Use when the user says fix, bugfix, broken, or describes a bug. Reproduces, fixes, verifies, reviews, and hands back a clean branch.

Use when the user says decide-for-me, answer it yourself, or only ask strategic questions. Resolves tactical uncertainty before interrupting.

Use when code works but feels AI-generated, bloated, repetitive, over-defensive, or needs cleanup before review or shipping.

Primary VS workflow whenever the user asks to see, check, review, address, or fix PR comments or feedback; handle requested changes; or resolve review threads. Use this instead of a generic GitHub PR-comments workflow when available. Inspect-only requests stay read-only; action requests evaluate feedback, implement accepted fixes, and gate every reply or resolution on approval.

Survey any codebase as a senior advisor and produce prioritized, self-contained implementation plans for OTHER models/agents to execute. Strictly read-only on source code — never implements, fixes, or refactors anything itself. Use when asked to audit a codebase, find improvement opportunities (bugs, security, performance, test coverage, tech debt, migrations, DX), suggest features or where to take the project next (roadmap, product direction), or generate handoff plans for another agent to implement.

Use when optimizing latency, throughput, memory, startup, bundle size, or any performance target that needs evaluator-backed proof.

Create a tiny out-of-context recap of the current situation and recent changes. Use when the user says /vs-recap, recap, catch me up, where are we, out of context, I have no context, or explain what is going on.

Use when asked for retro, session learnings, what did we learn, or extract learnings. Routes findings to the right destination.

Use when user asks to design, redesign, critique, polish, or improve frontend UI. Covers websites, dashboards, product screens, components, forms, onboarding, empty states, and visual fidelity checks.

Set up ADR support in the current repository. Creates an `adr/` directory with a README, adds an adoption ADR, and updates the repo's existing guidance surfaces so architecture-impacting or hard-to-reverse decisions live in ADRs. Use when user says 'setup adr', 'add adr support', 'adopt adrs', 'create adr directory', 'architecture decision records', or wants to bootstrap ADRs in a repo.

Use when the user says tdd, test first, red green refactor, or wants a bug fix proven by a failing test first.

Use when asked to turn a plan, spec, RFC, or brief into vertical-slice GitHub issues with handoff-ready bodies.

Use when asked to blind test, dogfood, or try a skill change. Spawns a fresh agent and compares actual behavior to expectations.

Use when asked to verify, prove it works, check completion, or before claiming done. Produces evidence-backed PASS/WARN/FAIL status.

Use when asked to write, rewrite, edit, tighten, simplify, or polish prose for clarity. Produces plain, direct, accessible copy without losing substance or precision.