@garrytan
Writes software, dreams.
148 skills3 repositories 168.7K
Set up and use gstack in Capy or Claude Code for planning, code review, debugging, and release work. It checks the local environment, handles installation or repairs when needed, and selects the workflow instructions for the task.
Plan a paid evaluation with predefined success criteria, comparison groups, decoy items, and stopping rules. Rehearse without spending, then run a three-item paid pilot and use its spending table to inform approval.
Create engineering retrospectives from Git history, covering delivered changes, work patterns, test code ratios, and frequently changed files. Review individual contributions and improvement opportunities, save metrics, and compare results with earlier periods.
Retrieve the top 30 stories from the Hacker News front page as JSON, a structured data format. Each entry includes its rank, title, link, points, and comment count for reviewing stories and their engagement.
Remove the directory editing boundary set by `/freeze` so work can continue across all directories without ending the session. The existing freeze hooks remain registered, and `/freeze` can be used again to restore a boundary.
Investigate software errors by tracing code, reviewing recent changes, and reproducing symptoms. Test a specific cause before making a minimal fix, then verify the result with regression tests and a structured debug report.
Explore a product or side-project idea through guided questions, challenge its assumptions, and compare implementation approaches. Produce a design document with a chosen direction and concrete next steps, adapting the discussion to business or personal goals.
Refresh gbrain's searchable code data to match the repository and update agent search guidance in CLAUDE.md. Use it to reindex a project or investigate searches that no longer find the expected code.
Review a development plan before implementation by challenging its premise, comparing alternative approaches, and examining design, security, testing, and operations. Choose whether to expand, maintain, or reduce scope, then receive a summary of agreed decisions.
Audit an existing test suite to find low-value or duplicate tests and code kept solely for those tests. Receive findings for review, with changes limited to batches you explicitly approve.
Prepare code changes for sharing by merging the base branch, running tests, and reviewing the diff. Update VERSION and CHANGELOG, then commit, push, and create a pull request, a proposal to merge the changes.
Turn a vague request into a precise specification through five phases and file it as a GitHub issue. Optionally start a Claude Code agent in a separate working directory to begin implementation.
Turn the most recent successful /scrape workflow into a reusable browser skill saved on disk. It builds a script, tests, and sample test data from the conversation, checks them, and asks before committing the result.
Turn an English description or mermaid source into a diagram with source files, an Excalidraw editing file, and SVG and PNG images. Rendering works offline, and flowcharts and sequence diagrams support editing individual shapes and arrows.
Set up gbrain so a coding agent can call it. Install its command-line tool, initialize a local PGLite or Supabase brain, register MCP, the agent connection mechanism, and record trust policies for each remote.
Update gstack by detecting how it is installed, refreshing registered installations, and handling project copies and version migrations. Receive a summary of user-facing changes, along with installation-specific recovery information if an upgrade fails.
Measure web page loading times, rendering milestones, and resource sizes in a real browser. Save baseline measurements, compare results before and after code changes, and inspect historical trends to spot performance regressions.
Clean up an iOS app after /ios-qa by removing the DebugBridge package and its debug wiring. The cleanup covers StateServer, DebugOverlay, generated accessor code, and hooks added to the app.
Enable warnings for destructive commands and restrict file edits to a chosen directory. It combines /careful and /freeze to help control work on live systems, while blocking certain catastrophic command patterns outright.
Monitor a live app after deployment for console errors, slower performance, and page failures. Periodic screenshots and comparisons with pre-deployment baselines help identify anomalies and check whether the release is behaving as expected.
Review proposed code changes before merging them. Compare changes with the base branch to find structural issues, including unsafe SQL, misplaced trust in large language model inputs, and unintended effects within conditional code.
Review plans for APIs and developer tools through the eyes of their users. Explore developer needs, compare competitors, and identify onboarding friction, with modes for competitive differentiation, detailed refinement, or critical gaps.
Use OpenAI Codex CLI to review code changes, challenge an implementation by looking for ways it could fail, or discuss technical questions. Three modes provide an independent opinion, with follow-up questions supported in ongoing consultations.
Turn approved mockups, design plans, or written descriptions into HTML and CSS using Pretext. Build dynamic layouts where text reflows and heights are calculated, choosing Pretext patterns suited to the design.
Is this your account? Link GitHub to bring it onto your creator page (Japanese)