blitz

Plugin blitz from marketplace blitz in lasswellt/blitz-cc. Commands read from the repo's manifest — 1 month ago.

Skills (20)

• audit

  Runs a 5-pillar recall audit (architecture, performance, security, maintainability, robustness) with paired agents plus registry checks, then writes docs/plans/audit-<date>/ as a paused task plan. Use for 'audit codebase', 'find tech debt', 'security audit', 'full code review', or before a release.

• browse

  Tests and crawls a running app through Playwright MCP: visits pages, captures console errors, failed requests and screenshots, classifies Critical/Error/Warning, optionally fixes sources; loop-safe, one page per tick. Use for 'test pages', 'smoke test', 'crawl site', 'click through the app'.

• check

  Runs the quality gate on a diff, a plan, or the repo: tsc/lint/tests/build, registry checks, task verify[], critic review; writes check-report.md. Use for 'check this', 'review my changes', 'run the gates', 'is this done', 'is this mergeable', 'check wiring/completeness/framework/design/security'.

• dep-health

  Audits npm dependencies for known vulnerabilities, outdated versions and license compliance. Modes: audit (read-only scan), upgrade (interactive bumps), report (full markdown report). Use for 'check deps', 'audit dependencies', 'security vulnerabilities', 'outdated packages', 'license check', weekly

• doc-gen

  Generates API docs, component docs and Mermaid architecture diagrams from source code into docs/generated/. Modes: api, components, architecture, full (parallel agents). Use for 'generate docs', 'API documentation', 'component docs', 'architecture diagram', or when source is ahead of docs/.

• doctor

  Checks the plugin install and the project's setup, fixes drift, and writes helper files: hooks, session state, CLAUDE.md conflicts, worktree.baseRef, loop.md, REVIEW.md, CI workflow, v2 migration. Use for 'is blitz healthy', 'set up blitz', 'migrate from v2', 'hooks misfiring'.

• learn

  Captures what a plan taught into docs/solutions/<slug>.md from progress.md rulings, the check report, and Task: commits, so plan reads it next time. Use for 'what did we learn', 'capture this', 'write this down', after a hard fix, or with --from-diff for an ad hoc fix outside a plan. Idempotent.

• migrate

  Migrates frameworks, libraries and tooling one major version at a time: researches breaking changes, plans atomic steps, verifies type-check, tests and build after each, aborts on repeated failure; --resume continues an interrupted run. Use for 'migrate to', 'upgrade to Vue 3', 'Vuex to Pinia', 'Nux

• next

  Recommends or, with --loop, dispatches the next unit of work from docs/plans/*/tasks.json and the inbox. Use for 'what next', 'where are we', 'run the loop' or any autonomous tick. One tick reconciles state, runs one row (build, check, learn+archive), commits with a Task: trailer, exits.

• onboard

  Sets a project up for blitz: scaffolds a new Vue/Nuxt/Firebase app or package, or maps an existing repo into CODEBASE-MAP.md. Use for 'new project', 'scaffold', 'set up a Nuxt app', 'add a package', 'I inherited this repo', 'map the codebase', or when no CODEBASE-MAP.md exists.

• perf-profile

  Profiles bundle size, runtime performance (Web Vitals) and Lighthouse scores for Vue/Nuxt apps and ranks optimizations (large deps, missing code splitting, render bottlenecks) by impact. Use for 'profile perf', 'lighthouse', 'bundle size', 'performance', 'why is this slow', 'optimize Vue/Nuxt'.

• plan

  Turns a goal into docs/plans/<slug>/{spec.md, plan.md, tasks.json}: brainstorm, research, tasks with executable verify checks. Use for 'plan X', 'spec this', 'break this down', 'what would it take', or before /blitz:build on anything larger than a one-sentence diff.

• refactor

  Refactors a file or module in small verified steps: snapshots the test baseline, changes one thing at a time, type-checks and tests after each, reverts on regression. Use for 'refactor', 'extract', 'simplify', 'decompose', 'rename', 'restructure', 'clean up'. Behavior changes go to /blitz:build.

• research

  Researches a topic with parallel agents (docs, web, codebase) and writes docs/research/<date>_<topic>.md with a Recommendation that /blitz:plan --from-research reads. Use for 'research X', 'compare options', 'evaluate Y', or 'how does this codebase do Y' (--codebase: read-only, file:line answers).

• sessions

  Use when asked what is running, who needs input, session status, leftover worktrees, or a session dashboard. Lists, triages, and prunes session records and platform worktrees. Modes: list, attention (HEARTBEAT_OK when empty), dashboard (--html), prune, worktrees (--apply removes merged, non-live).

• ship

  Releases finished work: checks the plan's gate, computes semver from commits, writes CHANGELOG, tags, publishes, archives the plan, runs learn. Slash-only. Use for 'ship it', 'cut a release', 'publish', 'release vX.Y.Z', 'tag this'. Refuses when any task is open or the check report is not PASS.

• test-gen

  Generates tests for a target file in the project's conventions (Vitest/Jest, AAA, factories): untested exports, edge cases, error paths; mocks only true externals; runs each test until it passes. Use for 'add tests', 'generate tests for', 'write tests', 'cover this file', 'test coverage'.

• todo

  Use when the user says 'todo: X', 'remember to X', 'add a todo', 'what's on my todo list', or when Claude surfaces a follow-up that should not become a stale TODO comment. Tracks ideas, follow-ups, and tech debt in docs/plans/BACKLOG.md (markdown checklist). Modes: add, list, check, resolve.

• ui-audit

  Audits a running app across pages and roles: builds a labeled value registry, asserts cross-page invariants (same field, same value; no role leaks), flags placeholders and flapping values, runs UI/UX heuristics. Read-only on the target app, loop-safe. Use for 'audit consistency', 'role leak', 'place

• ui-build

  Use when the user says 'build a page', 'create UI', 'add a form', 'build UI for X', 'add a screen', or 'extract design system' / 'build DESIGN.md'. Discovers the project's components, tokens, and conventions (extracting DESIGN.md when absent), then builds production-grade Vue 3 UI native to it.

Subagents (5)

• critic

  Fresh-context, read-only evaluator with two modes selected by the spawn prompt. `MODE: reject` tries to find ONE reason to reject a plan's diff (shortcuts, hallucinated APIs, ratchet regressions, deleted tests, --no-verify, mocked deps that should be real; reject-authority rows of /_shared/check-reg

• design-critic

  Design-quality vision critic. Reads screenshots of a rendered page and scores aesthetic fit, visual polish, prompt adherence, UX, and creative distinction against the project's DESIGN.md (or frontend-design heuristics if no DESIGN.md). Used by ui-build Phase 5.4 and the visual-iteration loop. Read-o

• dev

  Implements one task from docs/plans/<slug>/tasks.json. Role-neutral: the spawn prompt carries `ROLE: backend|frontend|infra|test` and inlines the matching skills/build/references/<role>.md conventions. Writes production-quality code with strict typing, structured errors, and real implementations (no

• research-critic

  Read-only adversarial reviewer for produced research docs in docs/research/. Probes every cited URL via WebFetch HEAD-equivalent and classifies each LIVE / DEAD / LIKELY_HALLUCINATED / UNKNOWN per Blitz's 4-way scheme (extends arxiv 2604.03173's 2-way hallucinated/non-resolving distinction). Verifie

• test-writer

  Test specialist for unit tests, integration tests, and E2E tests. Generates tests following AAA pattern with factory functions. Adapts to project's test framework (Vitest or Jest). The test-authoring contract a `dev` spawned by /blitz:build with `ROLE: test` reads: skills/build/references/test.md po

Hooks (30)

• PreCompact

  "${CLAUDE_PLUGIN_ROOT}"/hooks/scripts/pre-compact-snapshot.sh

• UserPromptExpansion

  "${CLAUDE_PLUGIN_ROOT}"/hooks/scripts/blitz-prompt-expansion.sh

• SessionStart

  "${CLAUDE_PLUGIN_ROOT}"/hooks/scripts/session-start.sh

• PostToolUse

  "${CLAUDE_PLUGIN_ROOT}"/hooks/scripts/post-edit-format.sh

• PostToolUse

  "${CLAUDE_PLUGIN_ROOT}"/hooks/scripts/post-edit-test.sh

• PostToolUse

  "${CLAUDE_PLUGIN_ROOT}"/hooks/scripts/skill-frontmatter-validate.sh --all

• PostToolUse

  "${CLAUDE_PLUGIN_ROOT}"/hooks/scripts/agent-frontmatter-validate.sh --all

• PostToolUse

  "${CLAUDE_PLUGIN_ROOT}"/hooks/scripts/post-edit-typecheck-block.sh

• PostToolBatch

  "${CLAUDE_PLUGIN_ROOT}"/hooks/scripts/heartbeat.sh

• StopFailure

  "${CLAUDE_PLUGIN_ROOT}"/hooks/scripts/stop-failure.sh

• SubagentStart

  ${CLAUDE_PLUGIN_ROOT}/hooks/scripts/subagent-context.sh

• WorktreeRemove

  "${CLAUDE_PLUGIN_ROOT}"/hooks/scripts/worktree-remove.sh

• PreToolUse

  ${CLAUDE_PLUGIN_ROOT}/hooks/scripts/kill-switch.sh

• PreToolUse

  ${CLAUDE_PLUGIN_ROOT}/hooks/scripts/tasks-guard.sh

• PreToolUse

  ${CLAUDE_PLUGIN_ROOT}/hooks/scripts/tasks-guard.sh

• PreToolUse

  "${CLAUDE_PLUGIN_ROOT}"/hooks/scripts/pre-edit-guard.sh

• PreToolUse

  "${CLAUDE_PLUGIN_ROOT}"/hooks/scripts/block-test-deletion.sh

• PreToolUse

  "${CLAUDE_PLUGIN_ROOT}"/hooks/scripts/block-as-any-insertion.sh

• PreToolUse

  "${CLAUDE_PLUGIN_ROOT}"/hooks/scripts/block-test-disabling.sh

• PreToolUse

  "${CLAUDE_PLUGIN_ROOT}"/hooks/scripts/pre-commit-validate.sh

• PreToolUse

  "${CLAUDE_PLUGIN_ROOT}"/hooks/scripts/block-no-verify.sh

• PreToolUse

  "${CLAUDE_PLUGIN_ROOT}"/hooks/scripts/block-destructive-git.sh

• PreToolUse

  "${CLAUDE_PLUGIN_ROOT}"/hooks/scripts/block-destructive-sql.sh

• PreToolUse

  "${CLAUDE_PLUGIN_ROOT}"/hooks/scripts/block-test-deletion.sh

• SessionEnd

  ${CLAUDE_PLUGIN_ROOT}/hooks/scripts/session-end.sh

• Stop

  ${CLAUDE_PLUGIN_ROOT}/hooks/scripts/stop-turn.sh

• Stop

  ${CLAUDE_PLUGIN_ROOT}/hooks/scripts/stop-gate.sh

• Notification

  ${CLAUDE_PLUGIN_ROOT}/hooks/scripts/notification-log.sh

• PermissionDenied

  ${CLAUDE_PLUGIN_ROOT}/hooks/scripts/permission-denied.sh

• ConfigChange

  ${CLAUDE_PLUGIN_ROOT}/hooks/scripts/config-change.sh

Language servers (4)

• typescript
• python
• rust
• go

Similar plugins