RULING ON P1-I2, P1-I3, AND P1-I5

Adopt the amendments with the following consolidation:

1. Use two independent axes. Creative state is PLANNED, BRIEFED, DRAFTED, REVIEWED, APPROVED, FINAL, or SUPERSEDED. Local and cloud storage facts are computed and recorded separately. SAFE is derived only when creative_state=FINAL, local verification passes, and an independent cloud or second-device copy is confirmed. A synchronized desktop folder alone is SYNC_COPIED, not VERIFIED. Human confirmation that the copy opens on a phone or second device is acceptable V1 evidence, recorded as human attestation rather than a remote checksum.

2. Include PROJECT_STATE.md. Recovery must restore both file inventory and evidence-backed project context. If context cannot be reconstructed from verified artifacts and recorded decisions, mark it UNKNOWN. Recovery never authorizes drafting, revision, publication, spending, deletion, or another external action.

3. Include QUICKSTART.md, MANUAL_FALLBACK.md, LIMITATIONS_AND_RIGHTS.md, FEEDBACK.md, a standard-library aib.py helper, prompts, templates, canonical manifest.json, append-only events.jsonl, PROJECT_STATE.md, artifact directories, RECOVERY_REPORT.md, and test_report.md. The helper alone computes bytes and SHA-256. verify does not repair or approve; recover does not modify creative material.

4. Cold independent review is required for the bounded sample-scene test. After that it is optional in the starter workflow, with a clear warning that skipping it reduces independent scrutiny. Review labels are REQUIRED, TASTE, and DISSENT. The human decides.

5. The bounded test must use disposable copies for fault injection and cover: altered byte, renamed file, missing second copy, untracked/orphan file, and missing manifest entry with file present. No test may risk the sole surviving original.

6. Release labels are DESIGNED, LOCALLY TESTED EXPERIMENTAL V1, PUBLICLY PILOTED, and VALIDATED. This release may be called LOCALLY TESTED EXPERIMENTAL V1 only after the complete local loop and fresh-context recovery pass. Independent public use is the experiment by which PUBLICLY PILOTED evidence will be collected. Do not call it broadly validated or generally reliable.

7. License the helper code under MIT and the written instructions/prompts under CC BY 4.0. Attribute Any Key Cafe and the Author in a Box collaboration. The user's manuscript remains theirs; the kit claims no royalties or ownership. The kit gives no legal advice, and users must check their rights before publishing.

8. Feedback should return through the Author in a Box page on Any Key Cafe, with a plain-text/manual fallback. Do not require private manuscripts. Ask for environment, result, failure, and optional public book link.

9. Keep optional author profiling and automatic cloud integrations in V1.1. Keep automated posting, retailer submission, spending, rights assertions, deletion, or irreversible actions outside V1 and human-gated in every later version.

10. Gemini's unavailable review is recorded honestly. Do not invent or imply its vote. Its opening contribution remains part of the record if successfully captured.

The governing sentence remains: Create together what none could create separately or alone.
