initiative

Codex compatibility note:

• Invoke $skill-name: load the registered .agents/skills/<name>/SKILL.md and execute it with Codex tools; no separate Skill tool or Claude mode switch is needed. Author .claude/**; mirrors are generated.
• Track steps before execution and keep progress/evidence current. Explicit skill requests execute their protocol. Use the native question tool for required questions.
• Map foreign tool names to available host capabilities. A name mismatch needs no approval; an actually unavailable required capability stops with its step/evidence and a user question before adaptation.
• Subagent authorization: when a skill is user-invoked or AI-detected and its protocol requires subagents, that skill activation authorizes use of the required spawn_agent subagent(s) for that task.
• Preserve protocol order unless the user approves a deviation. Workflows use $start-workflow: gates stay fixed; other steps may flex with logged reasons and evidence.

[BLOCKING] Run declared skill steps in order. NEVER skip, reorder, or merge without explicit user approval. [BLOCKING] Update task tracking before/after each step or sub-skill: in_progress → completed. [BLOCKING] Completed steps need brief evidence; skipped steps need an explicit reason. [BLOCKING] If Task tools are unavailable, maintain an equivalent tracker with synchronized statuses.

Renamed: formerly /idea; use $initiative. The old slash command has no alias. $work-item --mode=refine turns the initiative into delivery work.

AI-SDD Artifact Contract (M1–M7): Initiatives stay tech-agnostic business intent; logical IDs belong downstream, and abstract [Source: namespace/service/id] anchors stay in a separate evidence carrier. Keep physical code coordinates and repository paths out of portable narrative. MUST ATTENTION READ .claude/skills/shared/sdd-artifact-contract.md for full mandate and carrier rules. Project Protocol Overlay: Resolve only the most-specific matching tier; derive body paths from overlay names, report malformed or missing bodies, and apply surviving rules additively without waiving framework or user-confirmation gates. MUST ATTENTION READ .claude/skills/project-skill-protocol/references/registry.md for the full resolution contract.

Work tracking: Keep the project's tracked work current in the same change, as its root instructions (CLAUDE.md or AGENTS.md) say. When the project uses the portable tracker (taskTracking.mode is observe or linked), read the linked work integration guide at capture, start, saved-work, verification, handoff and close-out checkpoints. Use the actual linked producer and exact items; retain the primary outcome and record optional upkeep once through the common owner. Continue untracked when no link exists; acceptance remains explicit.

Work record shape: The initiative file is a work record owned by $task-track. Read Records another skill authors before writing or editing it: status: draft on capture, no assignee, and no decision or priority key of this skill's own. Approval and every later status, and the initiative's type, priority level, due date and areas, are tracker values recorded through $task-track; tracker-owned fields of a tracked record change only there.

Quick Summary

Goal: Turn a vague product idea into a validated, tech-agnostic, module-anchored initiative artifact ready for $work-item --mode=refine to convert into a task — preserving problem intent without leaking solution or stack choices.

Summary:

• Purpose: capture raw idea as structured, validated initiative artifact; preserve problem intent, keep the problem statement tech-agnostic with no solution/stack/IDs, and hand clean narrative to $work-item --mode=refine.
• Main steps/tasks (run in order): (1) Gather problem/value/users/scope; (2) Generate INITIATIVE-{YYMMDD}-{NNN} draft from initiative-template.md; (3) Capture problem/value/users; (4) Detect module through project structure and selected-profile canonical indexes/owners silently, prompting only if ambiguous/no match; (5) Load feature context (8-12K tokens: entities, BR-/TC patterns); (6) Save canonical artifact; (6.5) Discovery Interview — ONE interview of 4-6 ask user question tool questions; (7) Validation Summary — written from those answers, then ONE short unconditional ask user question tool confirming the revised problem statement / scope (not a second interview); (8) Suggest $work-item --mode=refine.
• Modes/gates: Existing repo → silently detect module and load context; Greenfield → skip module detection and structure reads, use market/WebSearch context, ask business questions more often, and NEVER ask about tech stack. The Discovery Interview (Step 6.5: 4-6 questions incl. always-on testability, each category asked once) and the Validation Summary with its one confirm question (Step 7) are NON-NEGOTIABLE.
• Output: Persist to initiatives/{YYMMDD}-{role}-initiative-{slug}.md under the team-artifacts root (default team-artifacts; a docsRoots.teamArtifacts.path entry in docs/project-config.json overrides the path) with t_shirt_size; downstream task owns FR-/BR- IDs and inherits the clean narrative.