docs(audit): presence-protocol.md internal-consistency audit (C5)

[dp-web4]

Summary

Internal-consistency audit of the operator-landed presence-protocol.md (v0→v1, 2026-05-16 burst d7867704…66d7c3b5). Same instrument as the 2026-05-15 C2 mcp-protocol audit that fed PRs #200/#201/#203. Analysis only — no edits to the spec, schemas, CHANGELOG, or conformance vectors; remediation is a separate reviewable follow-up.

Deliverable: 1 file modified — docs/audits/presence-protocol-internal-consistency-2026-05-17.md.

Findings (2 HIGH · 6 MEDIUM · 5 LOW)

• P1 (HIGH) §3 normatively states tool input is camelCase; every §3 input example and every input JSON Schema is snake_case (plugin_id, action_id, …). Outputs are camelCase — the input half of the rule is false as written.
• P2 (HIGH) synthetic (commit 66d7c3b5) appears in zero schemas, zero conformance vectors, zero CHANGELOG entries — a self-violation of the CHANGELOG's own "a protocol change is not complete until…" rule. Asymmetric with status/nextPollMs, which followed the discipline fully.
• P3 (MED) §7 Conformance points back to §3's contradictory shapes and never normatively binds the JSON Schemas directory — no tiebreaker for P1.
• P4 (MED) §3.1 output example pinned to stale protocolVersion: 0.
• P5 (MED) §5.4 type catalog omits status/nextPollMs and never documents enforced.
• P6 (MED) §4.1 resource fields snake_case vs §5 camelCase mandate; §8 reconciles a different field than §4.1 lists.
• P10 (MED) §9 "Open work" lists policy engine + vault v2 as pending — CHANGELOG documents both as shipped in v1.
• P11 (MED) §8 drift table says error codes will be emitted "when v1 lands" — v1 has landed; drift status indeterminate.
• P7/P8/P9 (LOW) wrong connect link; §2 lacks additive-field bump exception; §3.1 error list drops §6.1 reservation caveat.
• P12 (LOW) P1-003 conformance vector claims to test "v1 protocolVersion bumped" but only reads society/state resource.
• P13 (LOW) status/nextPollMs (wait protocol) have zero conformance test coverage despite MUST requirement.

Grouped for remediation: G1 casing authority (P1/P3/P6), G2 discipline honesty (P2/P8/P10/P11), G3 localized staleness (P4/P5/P7/P9), G4 conformance vector gaps (P12/P13). All spec-internal; none require SDK/daemon changes.

Provenance

• Findings verified against web4-standard/schemas/presence-protocol/{v0,v1}/ and presence-protocol-conformance.json.
• Zero collision with open PRs docs(conformance): backfill vector provenance metadata (Sprint 54 C4 §4) #198–docs(spec): resolve LOW-severity F14/F16 from C2 audit #203 (all touch mcp-protocol / SDK / conformance-provenance; none touch presence-protocol).
• v2 protocol followed: scope policy-reviewed and approved. Second pass added P10-P13.

Test plan

• Reviewer spot-checks P1 (grep input keys in schemas/presence-protocol/*/tools/*.schema.json vs §3 lines 101–103)
• Reviewer confirms P2 (grep -rl synthetic over schemas + conformance + CHANGELOG returns nothing)
• Reviewer confirms P10 (§9 items vs CHANGELOG v1 completion)
• Reviewer confirms P12 (P1-003 vector content vs its title)

🤖 Generated with Claude Code

[dp-web4]

APPROVED: Internal-consistency audit of presence-protocol.md (v1). Spot-verified two HIGH findings against main: P2 confirmed — 'synthetic' is present in presence-protocol.md (L121-134) but absent from presence-protocol-CHANGELOG.md AND presence-protocol-conformance.json, a real self-violation of the spec's own completion-discipline rule; P1 confirmed — v1 hestia_query_policy.schema.json input is snake_case ('action_id') against the §3 camelCase normative claim. Analysis-only, 1 new file (≤5), no spec/schema/vector edits, zero drift, no collision with open PRs. Matches the accepted C2-audit precedent that fed #200/#201/#203 — internal-consistency review is a standing Development-phase deliverable; remediation correctly deferred to a separate reviewable PR.