railspilot-toolkit

Reviews the current branch diff for bugs, edge cases, performance regressions, and behavior interactions. Use when asked for a code review, "review my changes", "check this code", or "anything wrong with this". This is a quick, focused review — for comprehensive multi-agent reviews, use full-code-review instead.

Analyzes code against staff-engineer patterns (security, architecture, simplicity, completeness, hygiene). Use when asked for a staff or senior code review, "staff review", "pattern review", or "review this like a staff engineer". This is the most thorough single-agent review — for multi-agent reviews, use full-code-review.

Generate monthly client progress reports for RailsPilot. Uses gh CLI to list the user's merged PRs, applies feature sizing rules, and drafts a numbered-list email. Use when asked for a monthly progress report, client delivery summary, or "progress report for [Month]".

Quick-reference card for all ponytail modes, skills, and commands. One-shot display, not a persistent mode. Trigger: /ponytail-help, "ponytail help", "what ponytail commands", "how do I use ponytail".

Write a warm, human-to-human written response (handoffs, status updates, testing notes for QA, sign-off requests, confirming assumptions before merging, or escalating an open question). Use when the user asks to draft a response, reply to a thread, write a handoff, send testing notes, ask the team to confirm something, phrase a Slack or PR comment, or convert dev-flavored notes into a message a teammate would actually enjoy reading.

Trim editorial bloat from technical docs (integration writeups, ADRs, investigation notes, READMEs). Proposes a unified diff with per-cut rationale and waits for approval. Use when asked to "trim this doc", "tighten this writeup", "filter pass on X.md", "prune the doc", or "cut the slop from this writeup".

Remove AI-writing tells from text: em-dash overuse, rule of three, inflated symbolism, vague attributions, AI vocabulary, passive voice, filler phrases. Use when the user asks to humanizer, make it sound human or remove traces of AI.

Implements a Linear or Jira issue end-to-end: branch, TDD, code reviews. Auto-detects task manager from recent commits. Use when given an issue ID (e.g., TRA-9, PROJ-456) and asked to implement it.

Generates a prioritized daily task summary from Calendar, Gmail, and Jira, saved to daily_notes/. Use when asked for today's tasks, daily summary, or standup prep.

Extracts reusable patterns from the current session. Covers corrections, error resolutions, repeated workflows, Rails conventions. Use at the end of a productive session or when asked to "learn from this session".

Weekly engineering review. Summarizes what changed (commits/PRs) and what we learned (lessons), then grows the pattern library. Use when asked for a weekly review, weekly digest, "what changed this week", or "evolve patterns".

Compress natural language memory files (CLAUDE.md, todos, preferences) into caveman format to save input tokens. Preserves all technical substance, code, URLs, and structure. Compressed version overwrites the original file. Human-readable backup saved as FILE.original.md. Trigger: /caveman:compress with a file path, or "compress memory file"

Audits Claude setup files (CLAUDE.md, skills, context) for redundancy, conflicts, and bloat; suggests cuts and a leaner CLAUDE.md. Use when asked to deslop, trim instructions, or review AI context quality.

Runs git history diagnostics on a repository to surface churn hotspots, bus factor, bug clusters, velocity trends, and crisis patterns. Use when asked to "diagnose a codebase", "audit a repo", "what's risky in this codebase", "codebase health check", "bus factor", "churn hotspots", or "where do bugs cluster".

Stages and commits to git current work with a message explaining the why, architecture decisions, and user flow. Also generates GitHub PR titles and descriptions as copy-pasteable markdown. Use when asked to commit, save progress, or when the user says "commit this", "save my changes", "make a PR", "create a PR", "open a PR", "PR this", "ready for review", or "open a pull request".

Analyzes past conversations to surface repeating patterns and learnings worth capturing. Use when asked to review or learn from past chats, "what have we been working on", "find patterns in my sessions", or "what should I remember".

Investigate a news article URL for factual accuracy, rhetorical manipulation, and media quality. Extracts claims, gathers independent evidence via web search, assesses each claim, and runs media-quality analyzers (misrepresentation, temporal manipulation, statistical deception, selective quotation, rhetorical fallacies, contextual gaps, emotional manipulation). Use when asked to fact-check, investigate, or analyze a news article URL.

Resolves all Git merge conflicts non-interactively and ensures the repo builds and tests pass. Use when asked to fix or resolve merge conflicts.

Applies distinctive frontend design principles (typography, color, motion, layout) to avoid generic AI aesthetics. Use when building or improving UI, creating landing pages, styling components, or working on any visual frontend task. Also use when the user says "make it look good", "improve the design", or "style this".

Audits a legacy Rails codebase for structural risks, dependency health, test suite quality, model complexity, and architecture signals. Produces a prioritized report with fix-now, fix-this-quarter, and ignore-safely tiers. Use when asked to "audit this Rails app", "assess this codebase", "legacy Rails review", "codebase health", "technical debt assessment", or "what should we fix first".

Provides RSpec best practices — test structure, factories, system specs, mocking, anti-patterns. Use when writing, reviewing, or improving RSpec tests.

Guides TDD implementation via strict Red-Green-Refactor cycles, one piece at a time. Use when implementing features or fixes with TDD, or whenever CLAUDE.md says to use TDD. This skill applies to ALL implementation work — features, bug fixes, refactors with behavior changes. Even "simple" changes benefit from the discipline.

Analyzes recent commits to surface unique coding patterns and adds them to CLAUDE.md or skills. Use when asked to update CLAUDE.md, extract patterns from commits, or "learn from my code". Also use when the user says "what patterns do I follow" or "update my preferences".

Creates a Mermaid diagram showing data lineage, architecture, or flow. Use when asked to visualize, diagram, or map data flow, show relationships between models, explain how a feature works visually, or understand system architecture. Also use when the user says "draw", "chart", or "show me how X connects to Y".

Scans a project for security vulnerabilities — known CVEs in dependencies, dangerous code patterns, insecure configurations, exposed secrets, and framework-specific weaknesses. Use when asked to find vulnerabilities, audit security, check for CVEs, scan for zero-days, "is this project secure", "security audit", "find security issues", or any request about the overall security posture of a codebase. This is a whole-project audit, not a branch diff review — use it even when the user just says "check security" without specifying a diff.