README FIRST — Ark RTL Review Witness Evaluation Kit v0.1

What Is This Packet?

This packet is a public evidence kit for Ark RTL Review Witness. It packages evidence documents, review-compression reports, historical repair-replay summaries, claim boundaries, and manifest lineage.

Ark turns a reference/candidate RTL pair into a bounded review packet with cone context, review-priority ranking, manifest lineage, and optional localized escalation evidence where a focused harness is available.

What Should I Open First?

• ARK RTL Review Witness Evidence Memo v1.0.html
• ARK RTL Review Witness Paid Evaluation Scope v0.1.html
• RTL_HISTORICAL_FIX_REPLAY_EXEC_SUMMARY_v0.1.html
• RTL_REVIEW_PRIORITY_COMPRESSION_v0.1.html

What Is Public Here?

The public packet shows the evidence format, review targets, limitation language, and manifest lineage. Additional technical implementation details are available during partner evaluation. Public controlled-mutation examples are described only as workflow evidence, not upstream vulnerability claims.

What Does The Evaluator Supply?

• A sanitized reference/golden RTL file or small RTL folder.
• A sanitized candidate/modified RTL file or small RTL folder.
• Optional context describing the change under review.
• Optional constraints or review questions.

What Does Ark Return?

• Replay packet JSON/Markdown.
• Review-compression summary.
• Isolated root targets and cone context where available.
• Review-priority targets.
• Manifest/checksum lineage.
• Claim-boundary notes and remaining human-review items.

What Is Not Claimed?

• Not full-chip verification.
• Not security signoff.
• Not guaranteed Trojan detection.
• Not proof that excluded regions are safe.
• Not a replacement for simulation, LEC, SAT, BMC, formal tools, or human review.
• Public RTL examples do not claim vulnerabilities in upstream projects.

5-Minute Review Path

1. 1. Open ARK RTL Review Witness Evidence Memo v1.0.html.
2. 2. Open RTL_REVIEW_PRIORITY_COMPRESSION_v0.1.html.
3. 3. Review RTL_HISTORICAL_FIX_REPLAY_EXEC_SUMMARY_v0.1.html.
4. 4. Open RTL_HISTORICAL_FIX_REPLAY_EXEC_SUMMARY_v0.1.html.
5. 5. Check packet manifest lineage and claim boundaries.

Historical Repair Replay Evidence

Added public retrospective repair-region evidence:

• RTL_HISTORICAL_FIX_REPLAY_EXEC_SUMMARY_v0.1.md / .html
• RTL_HISTORICAL_FIX_REPLAY_v0.1.md / .html / .json / .sha256
• RTL_PARTNER_HISTORICAL_FIX_VALIDATION_PROTOCOL_v0.1.md / .html
• RTL_REVIEW_COMPRESSION_OPERATIONAL_FAQ_v0.1.md / .html

Current smoke result: 15 public historical RTL repair cases across Ibex, CVA6, and OpenTitan; 13 cases produced ranked targets; top-3 exact-signal overlap was 12/13 versus deterministic random baseline mean top-3 rate 0.3219; median review-compression ratio was 349.3333x; median case runtime was 0.2013 seconds.

Boundary: this is historical repair-region overlap evidence, not proof of bug detection, formal signoff, or a claim that Ark would have found the original issue independently.