skills

Review the changes since a fixed point along three independent axes — Standards (does the code follow this repo's documented conventions and avoid known smells?), Spec (does it do what the issue/PRD asked?), and Architecture (does the change belong in the system and raise its evolvability — working with the grain of the existing seams, domain model, and layers to leave the architecture cheaper to change next — rather than lowering it?). Runs three parallel sub-agents that never share context and reports them side by side. The strongest architectural opportunity gets a concrete follow-up the operator can run. Use when the user wants to review a branch, a PR, or work-in-progress changes, or asks to "review since X".

Hand work to another agent through the mailbox — send to a named existing peer, or, when no name is given and the context calls for it, create a new agent wired to reply to you. Use when the operator says "dispatch to <name>" or "dispatch this to a new agent", optionally refining what to send.

Pure-filesystem inter-agent mailbox so any agent session can send a handoff document to another and pick up replies. Use when you need to set up your mailbox, send or reply to another agent, check for or wait on incoming messages, resolve another agent by name, take over / claim an existing mailbox identity (e.g. "take over Alice's mailbox"), or prune stale mailboxes from the registry. The fabric beneath the spawn, handback, and dispatch skills.

Return your work to the parent agent that spawned you by producing a handoff and sending it back through the mailbox. Use when you are a spawned/child session that has finished (or reached a checkpoint on) a side-quest and the operator wants the findings returned to the originating session.

Handoff the current conversation and open a new Claude instance in a fresh tmux window to continue the work. Use when the user wants to spin up a new agent session from the current one.

Generate a markdown changelog for the changes between a fixed point and HEAD, enriched with issue-tracker context. Use when asked for a changelog, release notes, or a summary of merged work over a range or time window. Pass through whatever the user names as the base (SHA, branch, tag, main, HEAD~5, or a date window); detect the issue tracker from the repo's environment; follow Keep a Changelog headings; update an existing CHANGELOG.md in place when one is present.

Ruthless token-minimization mode plus a "token police" prompt check. Tokens are water; you are dying of thirst. Drops filler, prefers single words and emojis, answers yes/no with ✅/❌, and flags bloated user prompts. Use when user says "desert mode", "token police", "save tokens", "be ruthless", "stop wasting tokens", or invokes /desert-mode.

Creates GitHub pull requests with repository templates, concise reviewer-focused descriptions, and Linear issue state updates. Use when the user asks to open, create, raise, or publish a PR for a GitHub-backed codebase, especially when branches are Linear issue-prefixed.

Generate chart images for markdown documents using the QuickChart API. Use when writing reports, READMEs, changelogs, dashboards, or any markdown output that needs an embedded chart, sparkline, gauge, or data visualisation. Covers bar, line, pie, radar, scatter, bubble, gauge, sparkline, sankey, progress bar, boxplot, candlestick, and mixed charts.

Creates git commits with a Gitmoji-prefixed Conventional Commit subject and changelog-aware messaging. Use when the user asks to create a commit, draft or refine commit text, choose a commit type or scope, map changes to Keep a Changelog categories, or add the best Gitmoji to a commit subject.

Compact the current conversation into a handoff document for another agent to pick up.