A collection of Claude Code skills, hooks, and scripts for structured plan review, codebase exploration, and AI-assisted development workflows.
| Skill |
Description |
| plan-ceo-review |
CEO/founder-mode strategic review — rethinks problems, challenges premises, expands or reduces scope across four modes |
| plan-design-review |
Designer-focused plan review — rates design completeness 0-10 across seven dimensions, generates mockups, identifies gaps |
| plan-eng-review |
Engineering manager-mode review — validates architecture, data flow, edge cases, test coverage, and performance |
| quick-plan-review |
Stress-tests plans through 20-40+ non-obvious questions covering implementation, UX, tradeoffs, and edge cases |
| office-hours |
YC-style forcing functions — startup mode (demand reality, wedge, specificity) and builder mode (design thinking brainstorm) |
| Skill |
Description |
| code-explorer |
Explore unfamiliar codebases — architecture, request flow tracing, onboarding guides, pattern search across local/GitHub/GitLab repos |
| generate-trd |
Generate Technical Requirements Documents through iterative questioning and codebase exploration |
| jira |
Fetch, create, update, transition, and search JIRA tickets |
| Skill |
Description |
| gemini-search |
Web search via Gemini — context-efficient alternative to built-in WebSearch |
| list-recent-models |
Fetch latest LLM model IDs, pricing, and metadata from Anthropic, Gemini, OpenAI, and OpenRouter |
| Skill |
Description |
| apple-hig-web |
Apply Apple Human Interface Guidelines to web — accessibility, color, typography, layout, navigation, motion |
| critique |
View syntax-highlighted diffs of uncommitted changes in a browser |
| Hook |
Description |
| session-context.sh |
Injects git branch, recent commits, TODOs, progress, and cross-project context into each session |
| session-end.sh |
Archives conversations, logs session stats, triggers async summarization via Gemini |
| pre-compact.sh |
Saves timestamped transcript archives before context compaction |
| log-tool-use.sh |
Logs all tool calls with timestamps and session IDs to ~/.claude/tool-usage.log |
| Hook |
Description |
| block-sensitive.sh |
Blocks modification of .env, credentials, secrets, keys, and certificates |
| block-websearch.sh |
Blocks built-in WebSearch, routes through gemini-search instead |
| notification-sound.sh |
Plays system sound on session events |
| stop-capture.sh |
Cleans up GIF recordings and screen captures on session end |
| Script |
Description |
| gemini-search.ts |
Bun/TypeScript backend for web search via Gemini API |
| list-models.ts |
Bun/TypeScript script querying Anthropic, Gemini, OpenAI, and OpenRouter for model metadata |
The skills/_shared/ directory contains reusable conventions used across skills:
- voice.md — Tone and writing standards (direct, concrete, no corporate speak)
- completeness-principle.md — Always recommend complete options over shortcuts
- ask-user-question-format.md — Structured format for decision-making questions
- review-outputs.md — Plan review reporting standards (one issue per question, file:line refs, lettered options)
- completion-status.md — Workflow completion protocol (DONE, BLOCKED, NEEDS_CONTEXT)
Planning and review skills (plan-ceo-review, plan-design-review, plan-eng-review, office-hours) are forked from Garry Tan's gstack skill framework, adapted to remove gstack-specific dependencies and follow Anthropic's skill authoring conventions.