darkroomengineering/darkroom

Parallel divergent ideation, framed generators, then a critic scores, clusters, and deepens the top 3. Triggers "brainstorm", "ideate", "widen the option space", or a fuzzy design or debugging question. Not for lookups or bugs with a known cause.

Repo audits, one mode or all (codebase, docs, process, perf, tests, threat-model, motion, SEO, debt). Triggers "audit the codebase", "full audit", "perf audit", "audit the tests", "threat model", "motion audit", "seo audit", "debt ledger". Single-page CWV goes to /lighthouse.

Autonomous mutate/score/keep optimization of a SKILL.md prompt. Triggers "autoresearch", "optimize skill", "tune this skill", "evolve this skill".

Build a feature, page, or integration end-to-end with a GO/NO-GO gate. Triggers "build a feature", "implement feature", "add feature", "build integration". Components go to /component, hooks to /hook.

Sync cc-settings with Claude Code and Codex upstreams, update a local install, or migrate a project's CLAUDE.md to AGENTS.md. Triggers "sync with claude code", "update cc-settings", "migrate to agents.md".

Save or restore a mid-task rollback point before a risky refactor, migration, or destructive edit. Triggers "checkpoint", "snapshot", "rollback to", "list checkpoints". End-of-session saves go to /handoff.

Claude-to-Codex bridge to delegate bulk mechanical work, get an independent cross-model diff review, or ask a read-only question. Triggers only when the user names Codex, as in "use codex", "have codex review", "cross-check with codex", "ask codex".

Create a UI component. Triggers "create component", "new component", "add component", or naming a component to build (Button, Header, Card).

Audit and prune rules, skills, and learnings to cut context bloat. Triggers "consolidate", "clean up rules", "what's redundant", or degraded agent performance from oversized instructions.

Build CONTEXT.md (domain glossary) and docs/adr/ through an interview. Triggers "domain language", "glossary", "ADR", or inconsistent terminology across the codebase.

Generate WCAG-checked type, color, spacing, and dark-mode tokens, or consolidate Tailwind v4 tokens without changing rendering. Triggers "design tokens", "type scale", "color palette", "theme setup", "consolidate tokens".

Initialize a new Darkroom project from the satus or novus starter. Triggers "new darkroom project", "satus", "novus", "scaffold from starter". Not the native /init, which writes CLAUDE.md.

Read-only codebase investigation through the Explore agent, including zooming out on an unfamiliar region. Triggers "how does X work", "where is X", "understand X", "where does this fit".

Debug and fix a reported defect, such as a failing build or tests, runtime or console errors, regressions, merge conflicts, red PR checks. Triggers "fix", "broken", "not working", "merge conflict", "fix CI".

Block Edit and Write outside one directory for the session to contain debugging or parallel agents. Triggers "freeze edits", "restrict edits to", "only edit this folder", "unfreeze".

Save state at the end of a session, or resume from the last one. Triggers "wrapping up", "running out of context", "done for today", "continue where we left off", "resume".

Turn proven session behavior into a skill, rule, profile, AGENTS.md diff, or team learning; artifacts seed /autoresearch evals. Triggers "harvest this workflow", "turn this session into a skill", "preserve this behavior".

Create a reusable React hook (useX). Triggers "create hook", "new hook", "custom hook", a useX-style name, or extracting logic out of a component.

Lighthouse audit of one page with an improvement loop until targets are met. Triggers "lighthouse", "page speed", "LCP", "CLS", "INP", "core web vitals". Repo-wide performance audits go to /audit performance.

Expert advice, risk review, or weighted comparison. Triggers "what should I", "advice on", "what could go wrong", "premortem", "compare approaches", "which is better", "tech selection".

Coordinate agents across plan, implement, test, and review, or fan out 3+ independent workstreams. Triggers "orchestrate", "parallel agents", "fan out", "split work", "overnight", "autonomous task".

Stress-test a plan through a product and founder lens before committing to it. Triggers "ceo review", "founder review", "product review", "should we even build this".

Clarify a vague scope through interview, then compile it into a PRD. Triggers "help me figure out", "define requirements", "PRD", "product spec", "feature spec".

Adversarial verification with three competing agents (issue-finder, disprover, judge). Triggers "double check", "are you sure", "poke holes", or a critical fix before production.

GitHub Issues as the plan of record; agents read and update them. Auto-invoke on issue-linked branches. Triggers "what's the plan", "project status", "sync with github", "close the issue".

Run the machine-verifiable gate (typecheck, test, lint, plus a screenshot for UI) so a diff is review-ready. Triggers "proof of work", "is this review-ready", "prove it is green".

Visual and accessibility QA with screenshot-first critique, contrast, touch targets, mockup-vs-implementation diff. Triggers "visual QA", "does this look right", "a11y check", or after component changes. Logic checks go to /poke-holes.

Behavior-preserving restructuring of code outside the current diff, such as extracting modules, renaming across files, or paying down debt. Triggers "refactor X", "restructure", "extract Y from Z". Just-changed code goes to /zero-tech-debt.

Weekly engineering retrospective with persistent metrics covering commit patterns, sessions, quality trends. Triggers "retro", "weekly review", "how was my week", "what did I ship", "show me my stats".

Review the local diff against Darkroom's checklist, or summarize feedback on the active PR. Triggers "review my changes", "check this diff", "summarize PR feedback". Native /code-review handles a diff or PR without the checklist.

Batch-review diffs from several agents with per-change re-entry cards and abridged reading diffs. Triggers "review batch", "what's pending review", "reading diff", or after fanning out several agents.

Promote a team-relevant learning to the shared team-knowledge repo after deduping against existing notes. Triggers "share this", "add to the knowledge base", or a gotcha, decision, or convention worth team-wide awareness.

Push, open a PR, and watch CI; land mode merges and cleans branches. Triggers "ship it", "create PR", "babysit CI", "land it", "fix CI and merge".

Product strategy advice on vision, positioning, and how architecture decisions serve them. Triggers "product strategy", "market positioning", "what should we build", "product direction".

Write or run tests, check coverage, or drive TDD. Triggers "write tests", "run tests", "coverage", "TDD", "red-green-refactor", or a pre-merge test pass.

Code analysis through TLDR (call graphs, impact, dataflow, semantic search) for far fewer tokens than reading files raw. Triggers "who calls X", "blast radius", or before a large file read or refactor.

First-pass, read-only review of a client or unfamiliar repo with ranked glaring issues. Triggers "triage this repo", "point out glaring issues", "first pass on this client repo".

Rewrite an accreted patch into its intended end state by removing unused compat layers, flags, fallbacks, and wrappers. Triggers "zero tech debt", "delete the compat layer", "kill the legacy path", "too many flags".