CJHwong

Spin up an agent team for a task with structured workflow. Provide a ticket reference, task description, or both.

Spin up a research agent team to explore a topic from multiple angles. No code changes.

Use this skill when the user wants to turn current uncommitted local changes into one or more commits: "commit my changes", "make a commit", "split this work into commits", "organize staged/unstaged changes", "write commit messages", or "use [why]/[how] commit format". Inspect git status and diffs, group files by logical change, propose a commit plan, and wait for approval before staging or committing. Do not use for rewriting existing commits, rebasing, pushing, creating PRs, or remote GitHub operations.

Use this skill when the user wants to clean up commits that already exist on the current branch: "squash commits", "rewrite history", "reword commit messages", "fold fixup commits", "autosquash", "clean up WIP commits", "organize branch history", or "prepare this branch for PR". Compare against the base branch, propose a safe rewrite plan, and wait for approval before rebasing or resetting. Do not use for uncommitted-only changes, creating new work commits, PR comments, CI checks, or GitHub remote operations.

Audit a document for unverified technical claims, verify attributions against real sources, fix errors, and provide honest critical review.

Use this skill whenever the user asks you to look at an image on their machine — extract text (OCR), describe a screenshot or photo, read a diagram or chart, or answer questions about visual content. Trigger even if the user doesn't explicitly say "vision" or "OCR" — phrases like "what's in this screenshot", "read this image", "transcribe this photo", "what does this diagram show", or "can you see what's in this picture" all qualify. Also trigger when the user references an image file (.png, .jpg, .webp, .gif) with information they need extracted.

Find stale .claude/projects/ folders whose source directories no longer exist on the filesystem, then clean them up (merge renamed, delete dead).

Generate a learnxinyminutes.com-style tutorial for any topic: a CLI command, programming language, library, framework, algorithm, tool, or concept. Trigger when the user asks to: - Create a learnxinyminutes tutorial for anything - Get a quick-start guide or quick overview of any topic - Understand how a command, tool, or concept works ("how does cat work", "explain grep") - Document a GitHub repo or local codebase they're exploring Trigger even without the word "learnxinyminutes" — "how does X work", "quick overview of X", "explain X to me", "quick-start guide for X" all qualify.

Use this skill when the user asks about Clean Architecture in Python — not generic theory, but the specific layer conventions (l1_entities, l2_use_cases, l3_interface_adapters, l4_frameworks_and_drivers), folder patterns, boundary interfaces, and .importlinter.ini contracts from CJHwong/py-clean-architecture-examples. Fetches live code from that repo as the authoritative reference. Also use when the user wants to scaffold a Python CLI, FastAPI, Django, or Celery project with CA layers.

Use this skill for Slack operations: listing channels, reading messages/threads, searching, posting messages, adding/removing reactions, managing user groups, or downloading attachments via the slack-cli binary. Trigger when the user wants to perform an action in Slack — "post to #channel," "react with thumbsup," "list channels," "who's in @backend," "download that file" — or retrieve raw Slack data. Do NOT use for answering questions about past Slack conversations (use gf-slack:slack-search instead).

Use when the user wants to read, search, or act on Slack from the terminal, even when they don't say the word "Slack." Triggers include a pasted Slack permalink (slack.com/archives/…); a #channel, thread, DM, or person referred to by name; or asks like "catch me up on #incidents", "what did Alice say in #eng", "search Slack for the deploy postmortem", "reply in that thread", "post this to #team", "react to her message", "DM Bob", or "who is @carol, is she online?". It covers reading channels, threads, and DMs, cross-channel search, people and channel lookups, reading files and canvases, and posting, editing, reacting, pinning, and scheduling messages. Prefer it over raw Slack Web API calls or a Slack MCP: each command returns one fully-resolved result in a single call.