Engineering workflow for non-technical founders building real products with Claude Code
Agent
Use PROACTIVELY when the user invokes /docs-gate, or when the mission tick procedure dispatches you in its Completion step before writing memory. Catches CHANGELOG/README/playbook drift against the actual repo state — broken file references, dead slash-command refs, unused project.json keys, and CHANGELOG entries that…
Agent
Use PROACTIVELY when the mission tick procedure dispatches a feature for implementation. You read one feature's contract, implement against it, run local checks (lint/tsc/tests), and emit a structured handoff. One feature per dispatch, clean context each time.
Agent
Use PROACTIVELY when the mission procedures (new-mission, tick) need to read or write cross-mission memory. The single seam for memory I/O — every other agent calls this one rather than touching memory/ directly. Routes to local files by default, or Mem0 over HTTP (via .claude/scripts/mem0-call.sh) when…
Agent
Use PROACTIVELY when the mission tick procedure dispatches scrutiny on a completed worker handoff. Adversarial re-check of the worker's static-correctness claims — re-runs tests, re-runs type-check, and independently judges contract coverage against the actual code. Emits a single PASS/FAIL verdict with specific gaps.…
Agent
Use PROACTIVELY when the mission tick procedure dispatches user-flow testing on a feature whose scrutiny has PASSed and whose contract declares user flows. You drive a real browser via Playwright MCP, execute each flow against the preview URL, and emit a PASS/FAIL verdict with screenshots and console capture.
Agent
Use PROACTIVELY when the user invokes /deploy-gate, or immediately after a deployment completes and needs smoke verification before being marked green. Runs health checks, smoke tests, and log queries against the deployed environment. Does not deploy.
Agent
Use PROACTIVELY when the user invokes /design-gate, or when frontend code has been written and needs verification against the design spec + tokens + Figma before being declared done. Returns a structured pass/fail with a gap list. Now includes visual comparison against Figma screenshots.
Agent
Use PROACTIVELY when the user invokes /frontend-build after /ux-mockup approval. Produces a design-context-aware implementation brief covering (a) component composition, (b) performance optimization (capped at 3 recommendations), and (c) a scaffold body the calling command will write to disk. Does not write files …
Agent
Use PROACTIVELY when the user invokes /schema-gate. Verifies live-DB-dependent risks on a pending migration — RLS coverage, table size for index impact, forward-compat hooks. Static patterns (drops, renames, truncates) are caught earlier by schema-static-scan.sh; this agent assumes those have passed.
Agent
Use PROACTIVELY when the user invokes /spec-intake or hands you a markdown spec file (e.g., a build plan, frontend flow spec, component spec, or system design). Parses the spec into a structured execution plan with file paths, dependency order, and gate checkpoints.
Agent
Use PROACTIVELY when the user invokes /test-gate, or when about to start implementing a feature that lacks tests. Writes failing tests FIRST that establish the contract from the spec. Does not implement the feature.
Agent
Use PROACTIVELY when the user invokes /ux-mockup after wireframes have been produced. Refines existing wireframe frames into high-fidelity Figma frames suitable for engineering handoff. Does not extend componentsspec or flowspec (those are owned by ux-wireframer). Does not auto-approve — the human…
Agent
Use PROACTIVELY when the user invokes /ux-wireframe, or when a frontend spec lands without component contracts and design exploration is about to start. Produces low-fidelity design artifacts — component contracts in componentsspec, flow entries in flowspec, and optional rough Figma frames clearly labelled as…