02 — Existing identity & approval channel — reuse-first

Goal: before staging anything new, find an existing mechanism that (a) records a real president/ai_council vote into apr_approvals and (b) authenticates the approver. Reuse-first, prefer existing identity substrate. No approval, no identity row, and no credential were created during this survey.

A. Approval-channel reuse (re-verified this turn)

candidate	records pres vote	records council vote	binds real id	no self-approve	no manual write	decision
dot-content-approve (DOT-018)	NO	NO	n/a	n/a	n/a	REJECT — content domain, not apr_approvals
dot-apr-propose/execute/health (309/310/311)	NO	NO	—	—	—	REJECT — lifecycle is propose→[GAP]→execute; no approve step
Directus API on apr_approvals	NO	NO	NO	NO	NO	REJECT — 0 directus_fields ⇒ raw write, approver is free text
DB functions touching apr_approvals (8)	NO	NO	—	—	—	REJECT — all READ votes; none INSERTs; no fn_record_approval
MCP query_pg/directus_create/write_file	NO	NO	—	—	NO	REJECT — agent-side manual write = fabrication/bypass
auto-approve (fn_auto_approve_add)	NO	NO	—	—	—	REJECT — neutered 2026-06-06 (writes review_note only)
historical 42 rows	NO	NO	NO	—	NO	REJECT — batch back-fill, not a repeatable channel

Result: no reusable lawful approval channel. The approve step (dot-apr-approve) is genuinely the missing primitive. (Confirms prior package apr-approval-channel-discovery-and-bootstrap, re-checked against live state.)

B. Identity-substrate reuse (the new question this macro forces)

Searched information_schema for identity/seat/credential/token/secret/council/president tables. Findings:

B1. qt001_authority_identity_registry ← closest existing identity substrate

Columns: identity_key, identity_class, reviewer_type, authority_lock(bool), evidence_path, active(bool), created_at.

• No credential/token/secret/hash column → it can declare an authority but cannot authenticate a caller.
• Rows: 1 — STAGE2.6A-FIX2-MACHINE-POLICY (class APPROVER, reviewer_type NULL, lock=true, active=true). Not a president, not an ai_council seat.
• Reuse verdict: reusable as an authorization declaration layer (is key X a valid seat?), but not as an authentication layer. A real substrate would extend it (or a sibling) with a token-hash binding — a new, governed mutation, which the owner deferred (HOLD).

B2. directus_users (13 accounts)

• nmhuyen@gmail.com (owner) — active, has token + password. The only plausible human-president holder.
• AI agent logins (gemini, agent.chatgpt, agent.claude_code_cli, agent.codex_cli, agent.gemini_cli, …) — all suspended, token-only. These are agent logins, not governance ai_council seats.
• No account named/identified as president (human governance seat) and no ai_council seat accounts.
• Reuse verdict: directus_users could back authentication (token/password), but there is no president seat and no council seats to bind to, and the AI accounts are suspended and operator-controlled (not independent).

B3. Historical apr_approvals approver identities (what a "seat" looked like)

president(human,14) · gemini(11) · gpt(11) · gemini-ai-council(2) · gpt-ai-council(2) · ai_council_1(1) · ai_council_2(1) — all free-text strings, zero credential binding. The quorum functions match these by approver ILIKE '%president%' / approver_type='ai_council' — i.e. a string test, trivially fakeable by anyone who can insert a row.

Reuse-first conclusion

• Approval channel: none reusable → dot-apr-approve is genuinely net-new (design in file 03).
• Identity substrate: an authorization registry exists (qt001_authority_identity_registry) and can be reused/extended, but no authenticated seat substrate exists, and a real one cannot be stood up in-session (file 04). Reuse-first is satisfied: we will not duplicate the registry; we extend it — but only once real seats exist and the owner consumes the warrant.