Fish Skills

Unified EOS operating system entry point. Context-aware routing to the right CEOS skill based on day of week, time in quarter, and current data state. Requires skinnyandbald/ceos skills to be installed.

Auto-detects your test framework, runs the full suite, diagnoses failures, and fixes them. Use after making code changes to verify tests pass, or when you encounter test failures you want diagnosed and fixed automatically.

This skill should be used when ending a session, switching contexts, or preparing for another Claude Code instance to continue work. It generates a structured handoff document capturing the current role, work state, git status, next steps, and notes so the next session can resume seamlessly. Also used to resume from a previous handoff document. Trigger words: handoff, hand off, session summary, wrap up session, pass the baton, context transfer, resume, pick up, continue from handoff, load handoff.

Generate professional PDF invoices from client billing data. Supports fixed-fee and hourly billing with auto-incrementing invoice IDs. Pass client name and line items as args, or invoke interactively.

One-command AI onboarding for any codebase. Generates AGENTS.md (project context) with CLAUDE.md symlink, STYLE_GUIDE.md (coding conventions), and Beads task tracking. Also configures machine-level settings (hooks, plugins). Skips files that already exist.

Copy generated text to the user's macOS clipboard using pbcopy. Use this skill PROACTIVELY whenever you've just generated content the user will paste somewhere else -- emails, messages, social posts, bios, copy, outreach sequences, or any standalone text block. Also use when the user says "copy", "clipboard", "copy that", "put that on my clipboard", or "I need to paste this". Don't wait for the user to ask -- if you wrote something they'll clearly paste elsewhere, offer to copy it.

Detects your project's full technology stack, then generates a comprehensive research prompt tailored to a specific topic like security, performance, or testing.

Reviews the feature you just built and adds missing test coverage. Focuses on behavior that matters — not coverage metrics. Use after completing a feature to identify untested code paths, edge cases, and risk areas.

Audit GitHub repo security settings, dependabot config, workflow hygiene, stale PRs/branches, and action pinning. Auto-fix drift with --fix. Use when you want to check repo health across an org or a single repo.

Builds a profile of your product (users, features, design system) then generates a research prompt for any UX or product design question.

Research any topic and produce a structured executive briefing optimized for rapid CEO decision-making.

Manages Git worktrees for isolated parallel development. Creates worktrees in .github/worktrees/ with symlinked .env files.

Socratic thinking partner that refines half-baked ideas into clear product or technical specifications through iterative questioning. Use when you have a vague concept, feature idea, or problem statement and need structured clarification before building.

Unified plan review — stack detection, Context7 staleness scan, multi-model counselors dispatch, and prioritized triage. Three modes: full pipeline (default), --dry-run (copyable prompt), --feedback (analyze external input).

Resolve PR review comments. Use when asked to fix PR feedback, given a PR URL, or on '/pr-resolution'.

Extract authentic marketing copy by interviewing the user as a buyer would, then analyze the output line-by-line. Use when the user says 'interview me for copy,' 'extract copy,' 'help me write copy by interviewing me,' 'copy interview,' 'what would I say about this offer,' or 'help me pitch this.' Combines buyer-perspective interviewing (inspired by Jason Adams' method) with line-by-line copy analysis and direct response principles.

Audit any project's CLAUDE.md (or AGENTS.md) against best practices and score its effectiveness. Use when reviewing a client's AI development setup, onboarding a new codebase, or improving your own CLAUDE.md. Triggers on requests like "audit my CLAUDE.md", "score my AI setup", "review my project context file", or "how good is my CLAUDE.md".

Run code simplification across entire codebase using parallel agents with automatic segmentation and coordination

Review code changes against STYLE_GUIDE.md before committing or creating PRs. Catches naming violations, import ordering issues, anti-pattern reintroduction, and convention drift. Run after /setup-ai or /generate-comprehensive-style-guide has created a style guide.

Triage, fix, and resolve bot code review comments (CodeRabbit, Gemini Code Assist) on PRs. Use when asked to resolve bot reviews, given a PR URL, or on '/resolve-bot-reviews'.

Progressively gather requirements through automated codebase discovery and yes/no questions, then generate a comprehensive requirements spec. Use when starting a new feature, planning a build, or when you need structured requirements before implementation.

Process meeting transcripts from Fireflies and Plaud with mandatory full transcript analysis, extract action items from ALL participants, create GitHub issues with smart repo routing, run deterministic verification gates, and generate EOS Level 10 Meeting summaries. When invoked bare (no arguments), scans both sources for unprocessed meetings and walks through them one by one. Use after team meetings or when the user mentions meetings, Fireflies, Plaud, L10, or action item extraction.

Review UI code for Web Interface Guidelines compliance. Use when asked to "review my UI", "check accessibility", "audit design", "review UX", or "check my site against best practices".

Enhanced scorecard entry with auto-pull from daily L10 meetings, Attio CRM, Google Calendar, Gmail, and filesystem. Walks through each metric one at a time with pre-populated suggestions for interactive confirmation.

Verify worktree plugin patches are intact after plugin updates. Checks compound-engineering and superpowers skills for Claude Code launch instructions.

Deep codebase analysis to generate or regenerate STYLE_GUIDE.md with full evidence citations. Use when /setup-ai's quick pass isn't thorough enough, when conventions have drifted, or after a major refactor. Produces a 17-section style guide citing specific files as evidence.