quirk

You MUST use this before any creative work - creating features, building components, adding functionality, or modifying behavior. Explores user intent, requirements and design before implementation.

Use to implement a multi-step task sequentially in one session when subagents aren't available — plans in context, then executes

Use when the user wants to deeply research a topic OR brainstorm/explore ideas WITHOUT building anything yet — produces a cited briefing or idea-landscape (an exploration doc), never a spec, plan, or implementation. Triggers on "do deep research on X", "research X", "let's brainstorm ideas around X", "explore X", "what are some ideas for X". NOT for "build/implement/add/fix X" — that is the brainstorming skill. Set wildness with `--wild 0.1–1.0`.

Use when executing implementation plans with independent tasks in the current session

Use when starting feature work that needs isolation from current workspace or before executing implementation plans - creates isolated git worktrees with smart directory selection and safety verification

The planning rubric the execution skills run in-context at the start of a run — turns the tech spec (tech.md) when present, else the logic spec / requirements, into a task breakdown (contracts, acceptance, parallelism) before code

The rubric the execution skills run in-context — only when the work warrants it — to author the code-anchored tech spec from the approved logic spec, before planning.

Use when dispatching the pi coding agent for a sub-task — code review, analysis, edits, or multi-agent orchestration. Triggers on "use pi", "run pi", or "pi <model>" (codex, terra, opus, sonnet, gemini, haiku, grok).

Use when releasing the quirk plugin - commits any pending changes, stamps a calendar-based (CalVer) version from today's date, writes the changelog, and pushes to origin autonomously. Triggers on "release quirk", "ship quirk", "bump and push quirk", or "cut a quirk release".

Use when creating, editing, or testing a Claude Code skill — writing SKILL.md and its frontmatter/description, choosing skill vs command vs subagent vs CLAUDE.md, fixing a skill that won't trigger, or deciding how strictly to test one. Covers activation (getting a skill to fire) and execution (getting it right once loaded).

Divergent-ideation subroutine for surfacing N non-obvious viable options when facing decision points. Explicitly invoked via /adhd or delegated from brainstorming step 6. Uses parallel agent dispatch (Diverge + Deepen) + inline Score/Cluster to generate structured option lists with critiques.

Use when a feature branch exceeds ~400 lines of diff, when a reviewer asks to "make this PR smaller", when planning a large feature so it ships in reviewable chunks, or when the user invokes /split-branch

Use when about to write hedging language like "pre-existing", "out of scope", "future work", "skipped for brevity", or "architectural concern" — instead, route the observation to a typed artifact (BUGS.md, DEFERRED.md, TEST_BACKLOG.md, proposals.md) or an ADR. Also use when the user invokes /quirk:artifacts:* commands or asks how to log a bug, defer a task, or record a decision.

Use when completing tasks, implementing major features, or before merging to verify work meets requirements

Use when encountering any bug, test failure, or unexpected behavior, before proposing fixes

Use when starting any conversation - establishes how to find and use skills, requiring Skill tool invocation before ANY response including clarifying questions