Tasks: the-loop should be compatible with Cursor

Phase 3 of 3. DAG of implementation tasks derived from the approved design.

Task list

• [x] 1. Research Cursor's plugin equivalent
  • Cursor 2.4 Agent Skills (SKILL.md standard), Cursor 2.5 plugins (.cursor-plugin/plugin.json + marketplace.json, auto-discovered skills/, commands/, rules/, agents/, hooks/), hook event model vs Claude's.
  • Depends on: none
  • Requirements: R1
  • Test: findings recorded in decision-015 (review-verified)
• [x] 2. Cursor manifests
  • .cursor-plugin/plugin.json (explicit skills//commands//rules/ paths, no hooks) and .cursor-plugin/marketplace.json, mirroring the .claude-plugin/ pair.
  • Depends on: 1
  • Requirements: R2
  • Test: CI JSON parse step includes both files
• [x] 3. Session-reminder rule
  • rules/the-loop.mdc (alwaysApply), guarded on .the-loop/config.yaml; note in hooks/hooks.json description that Cursor parity comes from the rule.
  • Depends on: 1
  • Requirements: R3
  • Test: uv run python -c "import json; json.load(open('hooks/hooks.json'))"
• [x] 4. Harness-neutral command prose
  • Annotate ${CLAUDE_PLUGIN_ROOT} references in init, upgrade-the-loop, create-design, create-tasks-plan, new-requirement to cover Cursor.
  • Depends on: 1
  • Requirements: R2
  • Test: npx markdownlint-cli2 "**/*.md"
• [x] 5. Docs + decision record
  • README (install/layout/commands note), skill + automation.md + workflow.md, architecture.md, roadmap.md (resolve the open question), decision-015 + index, this 3-phase spec + execution log.
  • Depends on: 2, 3, 4
  • Requirements: R1, R4
  • Test: npx markdownlint-cli2 "**/*.md"

Dependency graph (DAG)

1 → {2, 3, 4} → 5 (research gates everything; docs land last).

Checkpoints

After tasks 4 and 5: make check (ruff · pyright · schema validation · pytest · markdownlint) — the same command CI runs.