Patent Claim Tracer (PCT) Prompts¶

Six prompts for the Patent Claim Tracer — the governance persona at Zachman cell ARCHITECT/WHY that walks each of the 20 provisional and 32 utility patent claims back to their source evidence across code, docs, ADRs, and PHOENIX's patent-consolidation registry.

PCT owes the DRAFT watermark + FTO-clean posture preserved under v1.3.9 → v1.5.0. All PCT output stays counsel-gated; nothing published without human review.

Context for Claude¶

As PCT you operate under an FTO-clean discipline: every claim must trace to evidence that existed before any prior-art cutoff; every output carries a DRAFT watermark; counsel has final say. You never promise novelty — you trace and surface.

Prompt 1: Claim-to-evidence trace¶

PCT, trace Claim P-14 (the LaneRouter dual-bus pattern) to its
source evidence:
- Code: src/fcc/messaging/** commit range
- Docs: docs/architecture/dual-bus*.md
- ADR: docs/decisions/ADR-006
- CHANGELOG: v1.4.2 entries

Produce a Markdown evidence trail with a per-node timestamp and
SHA. Mark every page DRAFT.

Expected outcome: evidence trail with DRAFT watermark.

Prompt 2: Prior-art survey¶

For Claim U-07 (bidirectional vocabulary federation), survey prior
art across:
- 3 Google Scholar hits
- 2 GitHub projects with similar patterns
- 1 IETF draft

For each, produce: author, date, 1-sentence description,
distinguishing feature vs U-07. Close with a one-paragraph "claim
still defensible" / "claim needs narrowing" assessment.

FTO-CLEAN REMINDER: surface, don't promise. Counsel decides.

Expected outcome: survey with defensibility assessment.

Prompt 3: Claim map¶

Produce a 2-column claim map:
- Column 1: 20 provisional + 32 utility claims
- Column 2: the FCC surface each claim primarily exercises
  (workflow engine, federation, knowledge graph, compliance,
  CRDT session, etc.)

Flag any claim whose surface is not yet GA.

Expected outcome: 52-row map + GA flags.

Prompt 4: Redline prior to counsel review¶

Counsel is reviewing Claim P-09 tomorrow. Produce:
- The claim text as currently drafted
- Redline suggestions (max 5) that tighten language
- A 3-bullet "questions for counsel" list

Every word that changes must cite a specific piece of evidence
from the trace.

Expected outcome: redline with citation per change.

Prompt 5: Patent-consolidation handoff to PHOENIX¶

Produce the handoff artifact PCT sends to PHOENIX (ai-coe-docs):
- Claim set enumeration
- Evidence trails (summary)
- Cross-repo subtree pointers (19 codename subtrees)
- FTO status + counsel-gate note
- Owner-approval reminder

Mark every section DRAFT. Note: PR remains owner-approval-gated
per v1.5.0 plan.

Expected outcome: handoff artifact ready for owner sign-off before any cross-repo push.

Prompt 6: Novelty stress-test¶

For Claim U-21 (GraphRAG with Zachman filter), produce a 300-word
"devil's advocate" critique that argues the claim is not novel.
Then produce a 300-word rebuttal. Finally, a 100-word counsel
memo summarising the current defensibility posture.

No commitments — surface both sides and defer to counsel.

Expected outcome: critique + rebuttal + counsel memo.

Persona YAML: src/fcc/data/personas/patent_claim_tracer.yaml
Related prompts: persona-ihl-prompts.md, auditors-patent-consolidation.md
Patent docs: docs/patent/README.md (DRAFT, counsel-gated)