feat(M3): Shared-source generator for agent variants

skills/do-task/cursor/.generated-manifest.json

@@ -0,0 +1,19 @@
+{
+  "$schema": "https://ai-coding-skills.dev/schemas/generated-manifest/v1.json",
+  "generator": "scripts/generate-skills.mjs",
+  "generatedRoot": "skills/do-task/cursor",
+  "files": [
+    {
+      "path": "SKILL.md",
+      "kind": "file",
+      "mode": "644",
+      "sha256": "3a255d12f67be0eef3c45969befb23e3283a8a91f061053d939714d50e5f8cee"
+    },
+    {
+      "path": "templates/task-plan.md",
+      "kind": "file",
+      "mode": "644",
+      "sha256": "6af4c73434e932c1705ba13faf47be2152d61971f0b60ee9bece534d22493596"
+    }
+  ]
+}

skills/do-task/cursor/SKILL.md

@@ -3,6 +3,8 @@ name: do-task
 description: Execute a single user-supplied prompt end-to-end with two reviewer loops (plan review + implementation review) in Cursor Agent CLI. ALWAYS invoke when the user says `/do-task`, "do this task", "do task ...", "execute this task", or "make it so". Also invoke on the hint phrase "just do ...". Do NOT invoke on "implement this" (that phrase is reserved for implement-plan).
 ---
 
+<!-- ⚠️ GENERATED FILE – do not edit directly. Edit the canonical source in skills/do-task/_source/cursor/ and run `pnpm run sync:pi`. -->
+
 # Do Task (Cursor Agent CLI)
 
 Execute an ad-hoc user prompt end-to-end: parse → clarify → plan (with reviewer loop) → implement (TDD-first where applicable) → verify → implementation review loop → commit → optional push → notify.

skills/do-task/cursor/templates/task-plan.md

@@ -1,3 +1,4 @@
+<!-- ⚠️ GENERATED FILE – do not edit directly. Edit the canonical source in skills/do-task/_source/cursor/ and run `pnpm run sync:pi`. -->
 # Task Plan: [Short Title]
 
 > **Variant guardrail (Cursor):** Sub-skills (`brainstorming`, `test-driven-development`, `verification-before-completion`, `finishing-a-development-branch`, `using-git-worktrees`) MUST be invoked through workspace discovery from `.cursor/skills/superpowers/skills/<skill>/SKILL.md` or `~/.cursor/skills/superpowers/skills/<skill>/SKILL.md`. Reviewer invocations MUST use `--mode=ask --trust --output-format json`. `jq` is a hard prerequisite.