Handbook

Platform composition

STI is a leveraging extension on the Forge governed delivery control plane — not a standalone test product with its own run spine.

Updated

Responsibility matrix

Product STI uses it for STI does not own
forge-platform Schemas, ForgeRun attach points, workcell registration patterns Canonical platform policy
forge-lcdl Governed tasks (case design, triage, coverage) LLM transport in consumer repos
lmeta Campaign flows (generate → evaluate → run → evidence) Persistence, SQL, hook allowlists
forge-fleet Approved templates for pytest/Playwright/audit jobs Job scheduling policy
forge-lenses Evidence review, approval UI Test generation
Blueprints ISTQB methodology, versona-testing Runtime execution
Cursor Interactive authoring, local run triggers Unattended SoR

Critical rules (from platform boundaries)

  • Lenses is system of record for runs, approvals, evidence, and human review.
  • LCDL when a result affects a decision — use contracts or traceable governed flows.
  • Fleet exposes approved templates first; agents do not get raw execution power by default.
  • No agent workcell writes final state directly — return WorkcellResult linked to ForgeRun.

Attachment model

STI artifacts attach via ForgeRun extensions:

  • evidence.test_run_summary — test_run_summary.v1 payload
  • reasoning.lcdl_trace — optional LCDLTraceSummary from case-design tasks
  • execution.fleet_job — Fleet job correlation for containerized runs

Platform repo must not submodule forge-sti; cross-link by handbook URL and schema id only.

Cursor integration

See Cursor integration for rules, composer prompts, and versona-testing pairing.