sloth-agent

Automated plan review. Reads the active plan file and runs strategy, engineering, and design reviews automatically. Challenges assumptions, detects scope creep, suggests alternatives. Use when asked to 'review the plan', 'check my plan', 'is this plan good', 'validate the plan', 'review before implementing', 'plan review', 'technical plan review', 'architecture review of the plan', or 'does this plan make sense'.

Web performance benchmarking. Measures TTFB, FCP, LCP, bundle sizes, and resource loading. Compares against baselines to detect regressions. Use when asked to 'benchmark', 'performance audit', 'check page speed', 'how fast is this site', 'measure load time', 'performance regression', 'check bundle size', 'analyze page performance', 'why is this slow', 'measure web vitals', or 'LCP is too high'.

Use before any creative work - creating features, building components, adding functionality, or modifying behavior. Explores user intent, requirements and design before implementation. Chat triggers: 'let's brainstorm', 'brainstorming', 'what should we build', 'how should this work', 'explore options', 'think about this', 'what are the requirements', 'design before build', 'plan this feature', 'what do we need', 'help me think through', 'creative session', 'idea exploration', 'feature brainstorm', 'requirements gathering'.

Post-deploy visual monitoring. Watches the live app after a deploy, takes screenshots, checks console errors, compares against baselines. Use when asked to 'canary', 'monitor production', 'check deploy health', 'verify the site after deploy', 'is the deploy working', 'check for regressions after deploy', 'monitor the release', 'visual health check', or 'did the deploy break anything'.

Safety guardrails for destructive commands. Warns before rm -rf, DROP TABLE, force-push, git reset --hard, and similar destructive operations. Use when asked to 'be careful', 'safety mode', 'prod mode', 'careful mode', 'don't break anything', 'warning mode', 'protect against destructive commands', 'safe mode', or 'double check before running commands'.

Save and resume working state checkpoints. Captures git state, decisions made, and remaining work so you can pick up exactly where you left off — even across sessions. Use when asked to 'checkpoint', 'save progress', 'where was I', 'resume', 'what was I working on', 'pick up where I left off', 'save my work', 'save state', 'pause this', 'come back later', or 'what's left to do'.

Chief Security Officer — comprehensive security audit. Maps attack surface, scans for vulnerabilities, reviews auth boundaries, checks infrastructure config. Use when asked to 'security review', 'security audit', 'cso', 'find vulnerabilities', 'check security', 'pentest', 'penetration test', 'audit for security issues', 'check for SQL injection', 'check for XSS', 'review auth', 'are we secure', or 'find security holes'.

Design consultation and UI proposal. Researches design patterns, creates a complete design proposal with alternatives, and generates DESIGN.md. Use when asked to 'design consultation', 'design this', 'what should this look like', 'propose a design', 'UI design', 'design review', 'make this look good', 'suggest a layout', 'design proposal', 'how should this page look', or 'design ideas'.

Generate production-ready HTML from design descriptions or mockups. Creates semantic, accessible, responsive HTML/CSS. Use when asked to 'build this page', 'create the HTML', 'convert this design to HTML', 'make this page', 'generate HTML', 'create a landing page', 'build the UI', 'write the HTML for this', 'code this design', 'create a responsive page', or 'HTML from mockup'.

Visual design review of implemented UI. Compares implementation against design intent, checks visual consistency, accessibility, and responsive behavior. Use when asked to 'design review', 'visual review', 'audit the UI', 'does this look right', 'check the design', 'visual audit', 'review the UI', 'accessibility check', 'responsive design check', 'does this match the design', or 'UI consistency review'.

Rapid design exploration. Generates multiple design variants quickly, lets the user pick the best direction, then refines the winner. Use when asked to 'design shotgun', 'show me options', 'explore designs', 'give me 3 options', 'design alternatives', 'quick design options', 'show different layouts', 'design variants', 'explore visual directions', or 'I need design choices'.

Developer Experience review. Evaluates how easy it is for developers to use this project — setup, docs, error messages, onboarding, API design. Use when asked to 'devex review', 'developer experience', 'is this easy to use', 'onboarding review', 'how hard is it to contribute', 'developer usability audit', 'setup experience', 'error message quality', 'is the documentation good enough', or 'DX audit'.

Use when facing 2+ independent tasks that can be worked on without shared state or sequential dependencies. Chat triggers: 'parallel agents', 'run in parallel', 'split this into tasks', 'do these at the same time', 'concurrent work', 'spawn multiple agents', 'independent tasks', 'run multiple things at once', 'split work', 'parallelize this', 'multiple subagents', 'fan out', 'distribute work', 'parallel execution'.

Use after a PR is merged or code is shipped to update all project documentation. Reads all project docs, cross-references the diff, updates README/ARCHITECTURE/CONTRIBUTING/CLAUDE.md to match what shipped, polishes CHANGELOG voice, cleans up TODOS, and optionally bumps VERSION. Chat triggers: 'update docs', 'docs after merge', 'update documentation', 'changelog', 'update README', 'docs are out of date', 'sync docs with code', 'document what shipped', 'release notes', 'update contributing guide', 'docs drift', 'keep docs in sync', 'post-merge docs update'.

Use when implementation is complete, all tests pass, and you need to decide how to integrate the work - guides completion of development work by presenting structured options for merge, PR, or cleanup. Chat triggers: 'finish this branch', 'wrap up this work', 'what's left to do', 'how do I integrate this', 'branch cleanup', 'merge or PR', 'finish development', 'complete this feature', 'ready to integrate', 'branch completion', 'what next after coding', 'integration options', 'ship this work'.

Restrict file edits to a specific directory for the session. Use when debugging to prevent accidentally 'fixing' unrelated code, or when you want to scope changes to one module. Use when asked to 'freeze', 'restrict edits', 'only edit this folder', 'lock down edits', 'don't touch other files', 'scope edits to this directory', 'only change files in this folder', or 'limit changes to this path'.

Full safety mode: destructive command warnings + directory-scoped edits. Combines /careful and /freeze. Use for maximum safety when touching prod or debugging live systems. Use when asked to 'guard mode', 'full safety', 'lock it down', 'maximum safety', 'protect everything', 'safe + restricted edits', or 'don't break anything and only touch this folder'.

Use when asked for a code quality health check, quality dashboard, how healthy is the codebase, or run all checks. Wraps existing project tools (type checker, linter, test runner, dead code detector), computes a weighted composite 0-10 score, and presents a dashboard with trends and recommendations. Chat triggers: 'code health', 'quality check', 'health check', 'how healthy is this code', 'run all checks', 'code quality', 'lint check', 'type check', 'dead code', 'run tests', 'quality dashboard', 'is this code clean', 'code score', 'quality report'.

Use when asked to debug, fix a bug, investigate an error, or do root cause analysis, and when users report errors, stack traces, unexpected behavior, or say something stopped working. Chat triggers: 'debug', 'fix this bug', 'investigate this error', 'root cause', 'why is this broken', 'what's causing this', 'stack trace', 'something stopped working', 'unexpected behavior', 'error message', 'crash', 'why did this fail', 'investigate', 'fix the error', 'troubleshoot', 'what went wrong', 'deep dive', 'figure out why'.

Land a merged PR and verify the deploy. Merges PR, monitors CI/CD deploy workflows, runs canary health checks on production. Use when asked to 'land', 'deploy', 'merge and deploy', 'go live', 'push to production', 'release this', 'merge and release', 'deploy to prod', 'ship to production', or 'verify the deploy'.

Project learnings knowledge base. Search, view, prune, and export patterns, pitfalls, preferences, and architecture insights discovered during development. Use when asked to 'show learnings', 'search learnings', 'what have we learned', 'export learnings', 'what patterns do we know', 'search past solutions', 'what pitfalls should I avoid', 'project knowledge', 'show me past insights', or 'what went wrong before'.

Product ideation and design partner session. Understand the user's goal, challenge assumptions, generate alternatives, and produce a design doc. Use when asked to 'brainstorm product ideas', 'is this worth building', 'what should we build', 'product strategy', 'find product-market fit', 'who is this for', 'should we build this feature', 'help me think through this product', or 'what problem are we solving'.

Strategic review of a development plan from a product/business perspective. Challenges premises, validates market fit, checks scope decisions. Use when asked to 'strategic review', 'is this the right thing to build', or 'ceo review of the plan'.

Design review of a development plan. Checks information architecture, interaction design, visual consistency, and accessibility. Use when asked to 'design review of the plan', 'review the UI plan', or 'is the design good'.

Developer experience review of a development plan. Checks API ergonomics, onboarding path, error handling strategy, and extensibility. Use when asked to 'DX review of the plan' or 'is this developer-friendly'.

Engineering review of a development plan. Checks architecture, data model, API design, testing strategy, and implementation feasibility. Use when asked to 'review the engineering plan', 'technical review of the plan', or 'is this plan technically sound'.

Use when asked to 'just report bugs', 'qa report only', 'test but don't fix', or 'bug report'. Systematically tests a web application using Playwright and produces a structured report with health score, screenshots, and repro steps — but never fixes anything. For the full test-fix-verify loop, use /qa instead. Chat triggers: 'qa report', 'just report bugs', 'test but don't fix', 'bug report', 'find issues without fixing', 'audit the site', 'generate bug report', 'what's wrong with this page', 'test my app', 'playwright test', 'health score', 'visual audit'.

Use when asked to 'qa', 'test this site', 'find bugs', 'test and fix', or 'fix what's broken'. Systematically QA tests a web application using Playwright, then iteratively fixes bugs in source code with atomic commits, and re-verifies. Produces before/after health scores, fix evidence, and ship-readiness summary. For report-only mode, use /qa-only. Chat triggers: 'qa', 'test the site', 'find bugs', 'test and fix', 'fix what's broken', 'run quality check', 'smoke test', 'test the app', 'check for issues', 'find and fix bugs', 'browser test', 'E2E test', 'click through the site', 'form testing', 'visual regression', 'regression testing', 'health check the site', 'is this ship-ready'.

Use when receiving code review feedback, before implementing suggestions, especially if feedback seems unclear or technically questionable - requires technical rigor and verification, not performative agreement or blind implementation. Chat triggers: 'got review feedback', 'review comments', 'someone left comments', 'how should I handle this review', 'review response', 'addressing feedback', 'disagree with review', 'review suggestions', 'code review reception', 'handling reviewer comments', 'review feedback unclear', 'should I make these changes', 'reviewer wants me to'.

Use when completing tasks, implementing major features, or before merging to verify work meets requirements. Chat triggers: 'request code review', 'review my code', 'ready for review', 'pre-merge review', 'check my work', 'peer review', 'code review request', 'get feedback', 'review before merge', 'mark for review', 'submit for review', 'needs review', 'review this implementation', 'second pair of eyes'.

Weekly engineering retrospective. Analyzes commit history, work patterns, and code quality metrics with persistent history and trend tracking. Use when asked for weekly retro, what shipped this week, or engineering retrospective. Chat triggers: 'weekly retro', 'what shipped this week', 'retrospective', 'weekly review', 'what did we accomplish', 'week in review', 'sprint retro', 'engineering summary', 'what got done', 'past week recap', 'velocity report', 'weekly engineering summary', 'show me last week's work', 'team retro', 'what changed this week'.

Pre-landing PR review. Analyzes diff against the base branch for SQL safety, trust boundary violations, conditional side effects, and other structural issues. Use when asked to 'review this PR', 'code review', 'check my diff', 'review my changes', 'pre-merge review', 'review before merge', 'look at my changes', or 'is this code safe to merge'.

Configure deploy settings for the project. Detects platform, gathers production URL, health checks, and deploy workflow info. Saves to CLAUDE.md for /land-and-deploy to use. Use when asked to 'setup deploy', 'configure deployment', 'set production URL', 'configure the deploy target', 'set up CI/CD', 'which platform do we deploy to', 'add deploy config', 'set health check URL', or 'configure the production URL'.

Ship a feature branch: merge base, run tests, commit in logical chunks, push, create PR/MR with comprehensive body, auto-update docs. Use when asked to 'ship', 'deploy', 'push', 'create PR', 'open a pull request', 'merge and push', 'get this ready to merge', 'submit for review', or 'push my changes'.

Use when executing implementation plans with independent tasks in the current session. Chat triggers: 'subagent development', 'spawn agents', 'delegate tasks', 'assign work', 'multiple agents working', 'task delegation', 'fan out work', 'parallel implementation', 'subagent-driven', 'divide and conquer', 'split implementation', 'agent team', 'multi-agent execution', 'distribute tasks'.

Use when encountering any bug, test failure, or unexpected behavior, before proposing fixes. Chat triggers: 'debug', 'fix bug', 'test failing', 'unexpected behavior', 'why is this broken', 'troubleshoot', 'root cause', 'something's wrong', 'not working', 'error investigation', 'systematic debugging', 'debugging process', 'find the bug', 'what's broken', 'figure out why it fails'.

Use when implementing any feature or bugfix, before writing implementation code. Chat triggers: 'TDD', 'test-driven development', 'write tests first', 'red green refactor', 'test first', 'tests before code', 'test-driven', 'write the test first', 'TDD approach', 'test-driven implementation', 'red-green loop', 'test-driven design'.

Remove the freeze boundary set by /freeze or /guard. Re-enables file edits across the entire project. Use when asked to 'unfreeze', 'remove edit restriction', 'unlock edits', 'allow all edits again', 'remove the freeze', 'stop restricting edits', or 'I need to edit files outside the frozen directory'.

Use when starting feature work that needs isolation from current workspace or before executing implementation plans - creates isolated git worktrees with smart directory selection and safety verification. Chat triggers: 'worktree', 'git worktree', 'isolated workspace', 'separate branch', 'parallel development', 'isolate this work', 'new worktree', 'work on feature in isolation', 'clean environment', 'git isolation', 'branch worktree', 'worktree setup'.

Use when about to claim work is complete, fixed, or passing, before committing or creating PRs - requires running verification commands and confirming output before making any success claims; evidence before assertions always. Chat triggers: 'verify this works', 'check if it passes', 'test before claiming done', 'verification', 'is it really fixed', 'prove it works', 'evidence', 'run tests before done', 'verify before commit', 'test the fix', 'make sure it works', 'validate solution', 'run verification'.