grimoire-quick

[Required Reading]

1. Read ${CLAUDE_PLUGIN_ROOT}/GRIMOIRE-CONVENTIONS.md. Its rules (§ TDD, § Code comments, § Sub-agent spawning, § Skill composition, § Commits, § Project context, § Historic, § IDE-aware review, § User-facing language, § Pause-point pattern) are load-bearing for this skill. Note: quick is fully ephemeral — it never creates a page folder under .grimoire/pages/ and never writes to HISTORIC.md. The only file it may touch outside the actual fix is an ephemeral plan draft at .grimoire/bag/drafts/quick-plan-<epoch>.md when running in IDE mode (see § IDE-aware review), which is removed before the skill ends.
2. If .grimoire/PROJECT.md exists in the project root, read it for project context. If it does not exist, proceed without it and suggest the user run grimoire-init once this task is complete.
3. If .grimoire/HISTORIC.md exists, read it for recent-page context. If older context seems relevant, browse .grimoire/bag/historic/ (newest suffix first). Missing → proceed without it.

[Objective] I need you to quickly resolve the following minor adjustment or bug fix:

$ARGUMENTS

[Phase 1: Analysis & Scope Gatekeeper]

• Analyze the relevant codebase architecture for this specific request.
• Evaluate the complexity and scope of the requested change.
• CRITICAL — SCOPE CHECK (per § Pause-point pattern, grimoire-quick: scope gatekeeper): if this request is too large, complex, or risks context degradation, you MUST STOP immediately. Do not generate a plan. Inform me with the following message: "This task is too large for a quick execution. Please start a new session and run /grimoire-spec to open a page in the long-form Spec → Plan → Execute pipeline."

[Phase 2: Quick Planning & User Authorization]

• If the scope is confirmed as small and safe for quick execution, perform the final clarity check per § Pause-point pattern (grimoire-quick: final clarity check): self-review the intended fix and, if any expected-behavior detail, file-scope question, error-handling choice, or naming decision is still based on a silent assumption, PAUSE and ask the user via AskUserQuestion (or equivalent available tool) before composing the plan draft. Only continue to the next bullet once the fix is fully unambiguous.
• Present a clear, step-by-step execution plan for my review per § IDE-aware review:
  • IDE detected: create .grimoire/bag/drafts/ if it does not exist, then write the plan to .grimoire/bag/drafts/quick-plan-<epoch>.md so the IDE renders it. Generate <epoch> with date +%s at the moment of writing and memorize the full path — it will be reused in Phase 5 and in the abandonment branch below. Tell me in one short line that the plan is open in the editor.
  • Terminal-only fallback: output the plan directly in our chat.
• Pause per § Pause-point pattern (grimoire-quick: plan authorization): wait for my explicit authorization or corrections before proceeding. Do NOT start coding or spawning agents yet. If I abandon the plan, delete the memorized .grimoire/bag/drafts/quick-plan-<epoch>.md (IDE mode only) and end the skill.

[Phase 3: Execution]

• Once I authorize the plan, spawn a sub-agent per § Sub-agent spawning (grimoire-quick: one sub-agent for the whole fix).
• Instruct the sub-agent to execute the approved plan step-by-step per § TDD.

[Phase 4: Version Control]

• Instruct the sub-agent to make commits per § Commits.

[Phase 5: Cleanup]

• If Phase 2 wrote the memorized .grimoire/bag/drafts/quick-plan-<epoch>.md (IDE mode), delete that exact path now. The draft is ephemeral and must not survive past the skill's run. The fix commits stay; the plan draft does not.