Skip to content

Security: genforAI/pgsa-harness

Security

SECURITY.md

Security Policy

PGSA Harness Core stores project-coherence artifacts in the repository. Treat those artifacts as project state.

Do not store secrets

Do not put secrets, API keys, private tokens, customer data, credentials, private model outputs, or proprietary internal logs into PGSA artifacts.

Potentially sensitive files include:

  • pgsa/contracts/
  • pgsa/state/
  • pgsa/merge_proposals/
  • pgsa/reviews/
  • pgsa/integration/
  • pgsa/ledger/
  • pgsa/reports/
  • pgsa/gates/
  • pgsa/runtime/
  • pgsa/runtime/capability_requests/
  • pgsa/skills/
  • pgsa/evidence/
  • pgsa/factory/
  • pgsa/scenarios/
  • pgsa/audits/

Advanced artifact folders can contain operational assumptions, capability requests, runtime evidence references, review routing, or audit hypotheses. They should be reviewed before sharing a repository or report bundle.

External agent runs

A run without captured command, transcript, diffs, tests, versions, timing, and evaluator metadata must be labeled as a protocol run or dry run, not a real product run.

Reporting issues

Open a GitHub issue with a minimal reproduction and avoid sharing secrets or private project data.

There aren't any published security advisories