Handbook
Cursor integration
Cursor is the interactive control surface for STI (L0 autonomy). Runtime campaigns under ForgeRun may use cursor_cli as a governed workcell; IDE sessions author artifacts without replacing Lenses as system of record.
Updated
Surfaces
| Surface | Role for STI |
|---|---|
.cursor/rules/forge-sti-handbook.mdc |
Handbook sync + doc edit obligations |
versona-testing |
Discipline lens for strategy review and §5 evidence packs |
.cursor/composer/ (this repo) |
Operator prompts: author cases, run campaigns |
| Future Skills | /author-sti-cases, /run-sti-campaign (Blueprints templates) |
| Defect workflow | forge-defect-triage-rca-test-impact.prompt.md → STI case generation |
Natural loop in Cursor
- Human invokes composer prompt or skill with acceptance criteria / defect / risk input.
- Agent emits
test_case.v1/test_suite.v1JSON (or extends existing suite). - Human reviews; optional versona-testing session for formal gate.
- Run prompt binds suite to pytest/Playwright/Fleet; captures
test_run_summary.v1. - Evidence attaches to ForgeRun or local
evidence/until Lenses ingest.
Composer prompts (in-repo)
| Prompt | Purpose |
|---|---|
.cursor/composer/01-author-test-cases.md |
Generate layered cases from AC or test-impact doc |
.cursor/composer/02-run-test-campaign.md |
Execute bound suite and return evidence shape |
These prompts are not published to the handbook (generator skips .cursor/).
Rules vs runtime
IDE .mdc rules guide authoring. lmeta flows govern unattended multi-step campaigns. Do not embed LCDL task bodies only in chat — call contracts or CLI wrappers.
Related
- Layered test artifacts
- Platform composition
- Forge Platform: Cursor rules and skills (when live)