export-cursor-toolkit

name

export-cursor-toolkit

description

Exports .cursor/skills, .cursor/rules, and .cursor/agents from this repo to another repository, stripping deployed service-specific content and seeding a minimal docs/ set. Uses Mustache placeholders, tiered rules, and interview-driven sanitization. Supports greenfield scaffold and merge-into-existing targets. Does not export hooks. Use when exporting the Cursor pipeline toolkit, bootstrapping a new project from deployed service tooling, sanitizing skills/rules for reuse, or consolidating docs to the minimal standing set.

Export Cursor Toolkit

Copy the spec-driven pipeline (skills 00–19, shared references, workflow-state-manager) to a target repository with deployed service-specific content removed and a minimal docs/ stub set.

Detailed inventories, strip lists, and doc merge map: reference.md

Preconditions

• Source: this repository's .cursor/ tree (and doc-planner/templates/ for doc stubs).
• Target path is known (local directory or git clone).
• User has chosen export mode and optional rule tiers (see §Intake).

Intake (AskQuestion)

Run before copying files:

1. Target path — absolute path to destination repo root.
2. Mode — greenfield (create .cursor/ + docs/ + workflow-state.yaml) or merge (overlay into existing repo; never overwrite without confirming conflicts).
3. Optional rule tiers — from reference.md §Rule tiers; default none unless user selects Modal, CORS/browser, strict-typing, CI-after-push, etc.
4. Placeholder values — collect Mustache bindings (minimum: PROJECT_NAME, REPO_URL; optional: ORG_NAME, DEPLOY_TARGET, PRIMARY_LANGUAGE).
5. Conflict policy (merge mode) — skip / overwrite / ask per file.

Standing vs ephemeral documents

Export these docs/ stubs (Mustache placeholders, no deployed service content):

File	Contents
spec.md	Architecture, roadmap phases, data section, glossary
feature-list.md	Feature table + per-feature acceptance bullets
user-journeys.md	Single-file journey index + details
test-plan.md	Test matrix + QA assurance gates
api-contract.md	API surface (when applicable)
dependency-inventory.md	Dependencies and licenses
risk-register.md	Risks and mitigations
deploy.md	Deploy checklist + integration + runbook (merged)
adr/README.md	ADR process + link to template

Do not export as standing docs — session/ephemeral content lives in workflow-state.yaml only:

• execution-plan.md, config-spec.md, research-brief.md, context-brief.md
• Audit/verification outputs: qa-report.md, e2e-report.md, deploy-report.md, etc.
• Session logs: hotfix-log.md, bug-reports/, service-health-reports/

Update exported doc-planner skill and doc-types.md to reflect this manifest.

Export workflow

Copy this checklist and track progress:

Export progress:
- [ ] Intake complete (target, mode, tiers, placeholders, conflicts)
- [ ] Copy pipeline skills 00–19 + shared refs
- [ ] Generalize domain skills (data-management, service-health)
- [ ] Copy workflow-state-manager agent only
- [ ] Copy tiered rules (core + selected optional)
- [ ] Sanitize all copied files (interview-driven)
- [ ] Write docs/ stubs from merged templates
- [ ] Write empty workflow-state.yaml template
- [ ] Run post-export validation checklist
- [ ] Present summary + unresolved [TBD] items to user

Step 1 — Copy inventory

Always copy (see reference.md §Copy inventory):

• .cursor/skills/00-context through 19-address-pr-review
• Shared skill files: pipeline-preamble.md, workflow-state-reference.md, workflow-state-agent-protocol.md, considerations.md, connectivity-gates.md, deployment-catalog.md, template-registry.md, build-planner/, doc-planner/, gather-context/, audit-docs/, audit-licenses/, verify-build/, build-executor/, build-planner/, deploy-verify/, bug-investigation/, github-projects/ (if user wants)
• .cursor/agents/workflow-state-manager.md only

Generalize in place (rewrite deployed service domain → generic templates):

• data-management/SKILL.md — corpus/DB staging → generic data-asset staging
• 15-service-health/SKILL.md — RAG smokes → generic deployed-service health tiers
• config-validator/SKILL.md — remove deployed service env var names; use METAR

Never copy by default (confirm via AskQuestion if user wants optional extras):

• .cursor/hooks/ and hooks.json
• .cursor/agents/ except workflow-state-manager.md
• modal-jobs-monorepo/, clone-repos/, modal-proxy-header/ (org-specific)
• deployed service-only rules (see reference.md §deployed service-only rules)

Rules layout in target:

.cursor/rules/
  core/          # always exported
  optional/      # user-selected tiers only

Flatten to .cursor/rules/*.mdc only if the target project already uses a flat layout and user explicitly requests it.

Step 2 — Interview-driven sanitization

For each copied file (skills, rules, agents, templates):

1. Read the file.
2. Apply reference.md §Mustache bindings for identity fields.
3. Delete deployed service-only sections (RAG apps, chat-rag, DigitalOcean operator specs, bilingual i18n EV cycles, deployed service env vars, feature IDs F1–F31 table content).
4. Replace with generic placeholders or empty template sections (## Feature Scope with {{FEATURE_TABLE}} or instructional comment).
5. When ambiguous (shared pattern vs deployed service-specific), use AskQuestion — cite the file and line range; do not guess.

Sanitization is agent-led, not fully scripted. Use scripts/grep_deployed-service_leftovers.sh only for verification, not as the primary strip mechanism.

Step 3 — Docs stubs

Generate from .cursor/skills/doc-planner/templates/ with merges per reference.md §Doc merge map:

• Build docs/deploy.md from deployment-integration + deploy-checklist + staging-runbook sections (new merged template in export if not present in source).
• Fold acceptance-criteria bullets into feature-list.md template.
• Fold QA gates into test-plan.md template.
• Fold roadmap, data-management, glossary into spec.md template.

Use Mustache placeholders throughout. Remove HTML comment instruction blocks before presenting as final stubs unless user wants generator comments retained.

Step 4 — workflow-state.yaml

Copy templates/workflow-state.yaml to the target repo root. Fill project.name and project.description from intake placeholders. All stages pending; empty git_history, artifacts, issue_log, decisions_log.

Step 5 — Post-export validation

Run the checklist in reference.md §Validation checklist. Then:

bash .cursor/skills/export-cursor-toolkit/scripts/grep_deployed-service_leftovers.sh "{{TARGET_REPO}}"

Fix or flag every hit. Do not declare export complete while critical hits remain.

Merge mode notes

• If target already has .cursor/skills/pipeline/, diff before overwrite; prefer merging shared reference updates into existing files when the user chose ask conflict policy.
• Never copy deployed service workflow-state.yaml session data — always the empty template.
• Never copy docs/ content from deployed service — only generate stubs.

Output summary

Present to the user:

1. Target path and mode
2. Files copied / generalized / excluded counts
3. Optional rule tiers included
4. Standing docs created
5. Validation results (pass / warnings / manual follow-ups)
6. List of [TBD] / {{PLACEHOLDER}} fields the target project must fill before pipeline stage 01

Examples

Greenfield:

User: Export the cursor toolkit to ~/projects/acme-api — new repo, include Modal optional rules.

→ Intake → copy inventory → sanitize with interviews → docs/ stubs + workflow-state.yaml
→ validate → summary with PROJECT_NAME=acme-api

Merge:

User: Merge sanitized skills into ../legacy-app without hooks.

→ merge mode, conflict policy ask → skip existing hooks → overlay skills/rules
→ flag 3 files with manual merge needed