jdelgadoperez

End-to-end open-a-PR flow — run the staff-eng pre-flight self-review, commit and push via ship, then open the PR with a clear summary. Use when you say "open a PR", "create the PR", "ship this feature and open a PR", or "ready for review" and want the whole flow, not just a commit.

Batch-convert HEIC images (e.g., iPhone slide screenshots) to compressed JPGs that fit under Read tool size limits. Use when the user has HEIC files that need to be reviewed visually by Claude or batch-processed for any downstream tool.

Writing voice reference template for AI writing assistants. Captures authentic communication style patterns for code reviews, Slack messages, and other written communication.

Anti-sycophancy skill

Use before opening a PR, requesting reviews, marking a draft ready, or pushing a substantive code change to a feature branch — the author's own staff-engineer self-review gate. Triggers on "open a PR", "create the PR", "ready for review", "request reviews", "mark ready", "un-draft", "ship it", "I'm done", "ready to merge", "push this feature".

Recall full context from a past Claude session. Use when the user needs detailed context from previous work — architecture decisions, implementation approaches, debugging sessions, or conversation threads.

Search past Claude chat history using the memory-bank vector DB. Use when the user asks about previous conversations, past solutions, or wants to find something discussed in an earlier session.

Git worktree workflow. Commands, naming conventions, rules, and example workflow. Use when creating worktrees, switching branches, or starting parallel work.

Retrieve and synthesize the full context of a past conversation from the local memory-bank vector DB (Claude Code, Claude Desktop, ChatGPT history). Use when the user wants the whole story of an earlier session — "pull up everything from when we built X", "what did we conclude in that long debugging session", "recap the auth migration work" — rather than a few matching lines.

Semantic search over the user's past AI chat history (Claude Code, Claude Desktop, ChatGPT) via the local memory-bank vector DB. Use when the user asks what they did before, how they solved something previously, or to recall earlier decisions, bugs, or context — e.g. "what was that Docker fix I did", "have we hit this error before", "how did I set up auth last time".

Use when asked to review, audit, or refresh documentation and/or a project's GitHub wiki for accuracy — to check that user-facing docs still match what the code actually does after features ship. Catches drift between docs/README/wiki and the real CLI, dashboard, and behavior.