platty-corpus-quality

Platty Corpus Quality

Use this for Platty development and regression checks, not normal project analysis.

Commands

platty corpus run-fixture --id <fixture-id> --stage <stage> --json
platty corpus batch-report --framework <framework> --stage <stage> --json
platty corpus compare --id <fixture-id> --stage <stage> --json
platty corpus gate-check --id <fixture-id> --stage <stage> --json
platty corpus next-candidate --json
platty corpus audit-queue --json
platty corpus self-improve-once --id <fixture-id> --stage <stage> --dry-run --json

self-improve-once requires --dry-run from the packaged CLI.

Use the installed global platty binary for corpus commands. If platty corpus ... returns UNKNOWN_COMMAND, stop and report that the global CLI may be stale or the command may not exist. Rebuild/reinstall the global CLI before continuing.

Handoff

Use the Platty handoff card. The State line should include fixture id, stage, and pass/fail/missing expected status. The Recommended next line should name the next corpus command, or say none when the result is only a report.

Stop Conditions

• A corpus command fails with UNKNOWN_CORPUS_COMMAND or UNKNOWN_COMMAND: stop and report; do not invent subcommands or flags.
• FIXTURE_NOT_FOUND: stop and confirm the fixture id with the user or a listing — do not guess fixture ids.
• self-improve-once fails with SELF_IMPROVE_EXECUTION_REQUIRES_DRY_RUN: rerun with --dry-run. Real execution is intentionally disabled in the packaged CLI — do not look for a bypass.
• gate-check reports FIXTURE_GATE_FAILED: report the gate failure as the result — do not loop self-improve-once trying to force a pass.