Update to Lucia

{
  "description": "Review Vera, Clara or Lucia source, model-context paths, external services, public explanations and run-report contracts before releasing a product workflow.",
  "included_files": [
    {
      "relative_path": "agents/openai.yaml",
      "size_in_bytes": 278
    },
    {
      "relative_path": "references/manifest-contract.md",
      "size_in_bytes": 9174
    },
    {
      "relative_path": "references/model-data-report-contract.md",
      "size_in_bytes": 19725
    },
    {
      "relative_path": "references/public-process-page-contract.md",
      "size_in_bytes": 3876
    },
    {
      "relative_path": "scripts/local_helper_environment.json",
      "size_in_bytes": 318
    },
    {
      "relative_path": "scripts/local_helper_environment.py",
      "size_in_bytes": 1055
    },
    {
      "relative_path": "scripts/validate_privacy_surfaces.py",
      "size_in_bytes": 57562
    }
  ],
  "name": "privacy-surface-review",
  "skill_md_contents": "---\nname: privacy-surface-review\ndescription: Review Vera, Clara or Lucia source, model-context paths, external services, public explanations and run-report contracts before releasing a product workflow.\n---\n\n# Privacy surface review\n\nThis developer workflow is identical in Vera, Clara and Lucia. It does not add a\ncustomer intake, a consent questionnaire or a professional process to usage\ncounts. Ordinary authorized model processing follows the selected host account.\nFor Vera work, apply `../vera/references/work-continuity.md`: resolve prerequisites\nupfront and preserve the admitted job and its progress. Missing or failed privacy\nreporting remains a separate issue and must not block the professional pipeline\nor delivery of its reviewable result.\n\n1. Resolve this installation's product root and read `privacy/registry.json`,\n   `components.json` and `privacy/runtime-profiles.json`. Inventory every exposed\n   skill, embedded component, primary process and service. Studio Archive is\n   infrastructure. A direct skill must not escape review merely because it has\n   no embedded component. A registration alias maps an existing process; it does\n   not create another rated task.\n2. Read the selected complete skill, its references, relevant scripts, schemas,\n   model-visible payload builders, connectors and native or hosted routes.\n   Resolve a product source inside its root; resolve components in `modules/`\n   when installed or in the sibling repository component. Follow cross-component\n   dependencies. Inspect actual emitted payloads and app-only metadata separately.\n3. Describe information that can enter the selected host model context, with its\n   purpose and source. Names, identifiers, full documents and professional facts\n   can be necessary. Local calculations do not imply a local-only conversation\n   or automatic anonymization. Do not infer a provider or account configuration\n   from an operating system or a product name. Antigravity uses the provider\n   actually selected in that host; its package alone does not identify the model.\n4. Record every additional boundary: public retrieval, hosted service, connector,\n   and send/publish action. Reuse shared `privacy/services/` records through\n   `service_ids`. Preserve actual recipients, payloads, returned data, activation,\n   access, retention limitations and executable controls. A route being optional\n   does not mean it ran. Prior authorization applies only to the same scope and\n   only when the host permits it; never override an action-time approval or denial.\n5. Review each supported host route separately. Availability of local Python,\n   native UI, connected files and saved state differs. Explicitly record missing\n   capabilities; never claim a local helper executed in a chat-only host. The\n   user's account controls training and retention; the plugin does not inspect\n   or enforce them. Do not invent a per-case account form.\n6. For every registered public page, author its localized `public_explanations`\n   in the governing workstream manifest. Explain that function's actual data\n   path in plain language. The generated page uses this exact reviewed copy and\n   contract revision, selected by product and page. Keep “Quali dati arrivano al\n   modello” (localized) as the final process section. If the review is incomplete,\n   retain an explicitly labelled localized placeholder; never borrow another\n   function's behavior. `not-relevant` requires an evidence-based explanation.\n7. Read `references/model-data-report-contract.md`. The reviewed contract says\n   what can happen; the saved report says what happened. Review actual phase\n   evidence, identify deviations and unknowns, and record the semantic comparison.\n   Never fill actual counts or phases from this manifest. A contract digest is\n   not proof of model-provider transmission or professional correctness. Old\n   reports and their signatures remain verifiable without today's installation.\n8. Update `privacy/workstreams/<id>.json` or `privacy/services/<id>.json` only\n   after inspecting the governed source. Refresh the selected fingerprint with\n   the commands below. The refresh performs hashing, not semantic approval.\n9. Validate the complete register, regenerate contract/public artifacts and run\n   their drift checks. Rebuild every product/host affected by a shared change.\n   Review actual packaged manifests and preserved source declarations; package\n   validation is not evidence that a remote hosted service was deployed or tested.\n10. Exercise a synthetic workflow with a fresh supported session, show the actual\n    readable report by default, reopen it without rerunning the task, and verify\n    the retained signed receipt. Record the exact host, version and remaining\n    acceptance limits. Do not label CLI-only evidence as host-conversation proof.\n\n```bash\npython skills/privacy-surface-review/scripts/validate_privacy_surfaces.py --refresh <workstream>\npython skills/privacy-surface-review/scripts/validate_privacy_surfaces.py --refresh-service <service>\npython skills/privacy-surface-review/scripts/validate_privacy_surfaces.py\n```\n\nRepository generation and release:\n\n```bash\npython scripts/sync_privacy_runtime.py --check\npython scripts/build_privacy_contracts.py --check\npython scripts/build_product_release.py --check\n```\n\n## Judgment boundary\n\nCode checks schemas, explicit registrations, file identities, source freshness,\ntranslations present, exact contract bindings and package bytes. These are\nmechanically verifiable. The host agent judges relevance, adequacy, actual data\nexposure and the accuracy of the explanation from inspected evidence. Do not\nimplement keyword privacy classifiers or automatically delete or anonymize\nprofessional facts. Empty external-boundary/security-control arrays are valid\nwhen supported by source. A policy statement is not an executable control.\n\nThis register is engineering evidence, not legal advice, a DPIA, a provider\naccount audit or a certification of GDPR compliance.\n\nRead `references/public-process-page-contract.md` for the final HTML section.\nHonor an existing user choice only where the host permits prior authorization;\nit never overrides action-time confirmation or a denied operation. Do not invent\nper-run consent for the already authorized technical reporting route.\n"
}