Ultra-compressed communication mode. Cuts output tokens 65% (measured) by speaking like nobacot while keeping full technical accuracy. Supports intensity levels: lite, full (default), ultra, wenyan-lite, wenyan-full, wenyan-ultra. Use when user says "nobacot mode", "talk like nobacot", "use nobacot", "less tokens", "be brief", or invokes /nobacot. Also auto-triggers when token efficiency is requested.
Invoke the dev-squad agent swarm for collaborative development. Full-stack app building with 11 agents (coordinator, architect, designer, backend, frontend, reviewer, qa-engineer, auditor, devops, git-ops, writer). Phase 3.5 DESIGN gate (designer) prevents AI-slop UI by producing design tokens + visual spec + component inventory + responsive spec BEFORE frontend codes. Phase 0 Step 2.5 SaaS-mode auto-detect triggers full SaaS scope (multi-tenancy, billing, drill-down dashboards via saas-patterns Part 1+2). Phase 5 has formal iteration loop with rollback. Phase 6 SHIP pre-seeds .claude/ docs in generated apps + 180s auto-reviewer wait. PreToolUse security hook blocks 9 dangerous code patterns. Supports feature development, database tasks, bug fixes, architecture changes, security audits, infrastructure work, runtime/stability/quality auditing, and SaaS-class application building.
Compress natural language memory/doc files (.dev-squad/memory.md, gotchas.md, record.md, CLAUDE.md, or any project markdown) into lakonik-style dense prose to cut the INPUT-token cost of re-reading them every time they're injected. Preserves code blocks, URLs, paths, commands, and structure exactly. Compressed version overwrites the original; a full backup is kept outside the repo so it survives cleanup. Trigger: "/dev-squad memory-compress <filepath>" or "compress this memory file" / "compress gotchas.md".
React component testing with React Testing Library, Vitest/Jest, MSW for network mocking, accessibility assertions with axe, and the decision boundary between component tests and Playwright/Cypress end-to-end runs. Use when writing or fixing tests for React components, hooks, or pages.
Turn git history into user-facing release notes — the missing step between "PRs merged" and "release exists". Reads commits since the last tag, groups them by impact, translates conventional-commit subjects into language a user (not a committer) understands, and drops internal noise. Use at Phase 6/7 SHIP before tagging a release, and whenever the user says "changelog", "release notes", "what changed", "what's in this release", or "generate the CHANGELOG".
Detect scope creep — the operational form of "surgical changes" (dev-squad Rule 3). Compares the actual diff against the declared goal and flags everything the change touched that the task did not ask for: unrelated refactors, drive-by reformatting, speculative features, adjacent "while I'm here" edits. Use during Phase 5 review and before opening a PR, or when the user says "intent drift", "scope creep", "did this stay in scope", "is the PR doing too much", or "compare the diff to the goal".
Verify test QUALITY, not just coverage — the operational form of "tests encode intent" (dev-squad Rule 9). Mutates the code under test and measures how many mutants the suite kills; a surviving mutant is a regression the tests cannot catch. Use during Phase 5 review/verification on non-trivial logic (money, security, auth, parsers, state machines, branching), whenever a PR adds tests, or when the user says "mutation test", "are these tests any good", "test quality", "coverage lies", or doubts whether tests would catch a regression.
Hunt for failures that never surface — the bugs no test and no error log will catch because the code swallows them. Finds empty catch blocks, errors logged but not raised, ignored return values, unhandled promise rejections, bare excepts, and discarded Go errors. Use during Phase 5 review and in QA failure analysis, whenever code touches error paths, or when the user says "silent failure", "swallowed error", "why didn't this throw", "it fails quietly", or "find hidden bugs".