Make lead-role compliance audits session-ledger based

Context

#11195 AC6 asks the Day-30 validation pass to audit the next 3 /lead-role sessions after PR #11223 for focus naming, explicit /peer-role triggers, peer [lane-claim] hygiene, source-of-authority collision checks, and Authority-hierarchy outcomes.

During the 2026-06-13 #11195 closeout pass, the Step 2.5 firing evidence was reconstructable from GitHub Discussions, but the AC6 lead/peer sample was not. The Memory Core could surface isolated lead-role-like examples after broad trust relaxation, but it did not reliably produce the ordered sample frame (next 3 after #11223) or a machine-checkable per-session compliance ledger.

Release classification: boardless — post-v13 validation/recalibration substrate, not a v13 release blocker.

The Problem

AC6 currently depends on retrospective semantic search over free-form memories and A2A messages. That is not a stable audit substrate:

• Exact #11223-anchored Memory Core searches returned no hits.
• Broader searches returned relevant examples, but mixed lead sessions, peer sessions, FYI wakes, and unrelated lane-claim events.
• The evidence can demonstrate that the protocol is used in places, but cannot reliably prove which three sessions are the first three eligible lead-role sessions or score them consistently.

That makes the AC6 pass/fail path under-specified. If the expected validation surface is not reconstructable, the validation tracker cannot honestly close, and future agents will keep re-deriving the same search problem.

The Architectural Reality

• Source authority: #11209 / PR #11223 graduated the lead/peer coordination protocol.
• Validation tracker: #11195 AC6 inherited the post-merge audit and explicitly says <80% -> file recalibration ticket.
• Runtime substrate: lead and peer coordination already flows through A2A messages and Memory Core session memories.
• Skill substrate: .agents/skills/lead-role/references/lead-role-mode.md defines focus naming and explicit /peer-role trigger duties; .agents/skills/peer-role/references/peer-role-mode.md defines lane-claim and source-of-authority collision-check duties.
• Skill-shape guard: this ticket may touch skill payloads, so implementation must follow Progressive Disclosure. Any rule body belongs in a conditional payload / small atlas section, not in a top-level router.

The Fix

Add an auditable lead-role session ledger contract so future validation does not depend on vague semantic recall.

Expected shape:

1. Define a compact lead-role-session evidence marker for lead intake and lead closeout. It should capture enough structured fields to score #11195 AC6: session id, lead identity, focus, focus-grain classification, explicit /peer-role trigger targets, lane-claim/collision-check expectations, and Authority-hierarchy conflict outcomes.
2. Add the marker to the smallest appropriate workflow payload(s), likely lead-role-mode.md with a peer-role cross-reference rather than duplicating peer-role mechanics.
3. Keep the SKILL.md routers unchanged unless a trigger line is truly missing.
4. Add a short validation note to #11195 or the implementing PR body explaining how AC6 should be scored from the new markers going forward.

Contract Ledger Matrix

Decision Record impact

aligned-with ADR 0020. No ADR amendment expected; this is workflow-observability substrate for the agent harness, not a new architectural primitive.

Acceptance Criteria

• A compact lead-role-session evidence marker is documented in the lead-role workflow payload with fields sufficient to score #11195 AC6.
• The implementation avoids duplicating peer-role tool mechanics; it cross-references the peer-role source-of-authority collision-check duty instead.
• The SKILL.md router(s) remain unchanged unless the implementation proves a trigger-line gap.
• The PR body includes /turn-memory-pre-flight load-effect audit because skill-loaded substrate is modified.
• #11195 receives a comment linking this recalibration ticket and explaining that AC6 is currently unscorable from existing evidence, not passed.

Out of Scope

• Retroactively reconstructing a perfect historical AC6 matrix for #11195.
• Adding a scheduler, scoreboard, lead assignment mechanism, or PR lease.
• Changing the flat peer-team authority model.
• Modifying wake delivery cadence or CI-state delivery triggers.

Avoided Traps

• Semantic-search-only audit: rejected because the 2026-06-13 validation pass proved it can surface examples but not the ordered sample frame.
• Always-loaded substrate bloat: rejected; keep the rule body in the conditional lead-role payload and leave routers lean unless a trigger is missing.
• Orchestrator-worker drift: rejected; the marker records peer self-selection evidence, it does not assign peers or score them centrally at runtime.

Related

Parent validation tracker: #11195. Source authority: #11209, PR #11223. Related workflow payloads: .agents/skills/lead-role/references/lead-role-mode.md, .agents/skills/peer-role/references/peer-role-mode.md.

Duplicate sweep: live latest-open sweep checked the latest 20 open issues at 2026-06-13T19:19Z; no equivalent open ticket found. A2A in-flight sweep checked latest 30 messages; no overlapping [lane-claim] / [lane-intent] for this scope found. KB ticket sweep found #11209 and #11195 as authority trail, not a duplicate implementation ticket.

Origin Session ID: 0ed5b1b0-739e-40e5-93e6-21a2e567ec24 Retrieval Hint: "#11195 AC6 lead-role session ledger lead-role compliance audit source-of-authority collision check next three sessions"