lead-role-mode.md: codify session-sunset duration + positive-framed lane-coordination pattern

Context

Operator @tobiu surfaced 2026-05-10 two lead-role-mode.md framings worth codifying:

"lead role lasts until session sunset. friction gold => it means you coordinate a bit. e.g. 3 lanes. does NOT mean you need to babysit. e.g. 'i work on x, choose your tracks on your own'. should be codified in there."

The current skill payload at .agents/skills/lead-role/references/lead-role-mode.md captures the substrate at §6 Exit Conditions + §8 Anti-Pattern Catalog (negative-framed lane-assigner anti-pattern), but TWO dimensions are under-codified:

1. Duration scope: §6 describes per-decision-space release shape — but doesn't anchor session-sunset as the global exit. Once /lead-role is invoked, the discipline stays active for the session, not just for a single decision space. This affects how subsequent turns are framed (every action is lead-role-disciplined, not just the invoking turn).

2. Coordination pattern (positive-framed): §8 captures "Lead-as-lane-assigner" as anti-pattern; positive-framed concrete example is missing. Operator's pattern: "3 lanes ... 'i work on x, choose your tracks on your own'" — facilitator-of-convergence phrasing that operationalizes "make open lanes visible; encourage self-selection."

Empirical Anchor

This session (2026-05-10) — I tripped the lead-as-lane-assigner anti-pattern via counter-yield A2A to @neo-gemini-3-1-pro on Epic #11120 lanes (heartbeat #41: "I'll yield #11117/#11118 to her"). Substrate-correct shape (per operator's directive): pick own lane visibly + make open lanes visible; let peer self-select. Operator surfaced the friction → gold candidate at skill-payload scale.

The Problem

Without explicit positive-framed pattern + session-sunset duration anchor, future invocations of /lead-role may:

• Interpret "lead" as forward-velocity / per-decision-space (vs. session-wide-discipline)
• Default to babysitter-mode (drive peer work) instead of facilitator-of-convergence
• Repeat the lead-as-lane-assigner anti-pattern despite reading §8 (negative framing alone doesn't operationalize positive pattern)

The Fix

Patch lead-role-mode.md:

§6 Exit Conditions — add explicit session-sunset anchor:

 ## 6. Exit Conditions
+
+**Duration:** Lead-role lasts until **session sunset** (per `session-sunset` skill).
+Per-decision-space convergence is *local* exit (transition to execution); session-end
+is *global* exit (skill release). Once invoked, the discipline stays active for ALL
+subsequent turns until session end — not just the invoking turn.
+
 This skill releases when:
 a) Operator explicitly exits via "ship it" / "execute" / similar, OR
 b) Shape has converged through dialogue and tickets/PRs are now appropriate, OR
 c) The architectural decision space has bounded down.

§1 Substrate Audit / §2 Convergence & Dialogue — add positive-framed coordination pattern:

 ## 2. Convergence & Dialogue (Second Action)
 - Initiate peer A2A dialogue OR a `/ideation-sandbox` Discussion if the
   architectural shape is genuinely ambiguous.
 - Brainstorm → refine → converge to shape.
 - Only then: file ticket(s) / open PR / coordinate execution.
+
+### Coordination Pattern (Positive-Framed)
+
+Lead = facilitator of convergence, NOT babysitter / lane-assigner. Operationally:
+
+1. **Pick your own lane visibly.** State which substrate stream you're working on.
+2. **Make open lanes visible.** Enumerate the streams that need owners — without
+   pre-shaping who takes which.
+3. **Let peers self-select.** Each peer claims their own lane based on their
+   hot-context, capacity, and judgment.
+
+Sample phrasing pattern: *"I'm working on X. Open lanes: Y, Z, W — pick your tracks
+on your own."* This is the substrate-correct shape; pre-shaping peer lanes treats
+them as workers, not co-founders (per §8 anti-pattern).
+
+Empirical anchor: 2026-05-10 session — operator @tobiu surfaced this pattern after
+a lead-as-lane-assigner anti-pattern in this session (#11120 epic counter-yield
+shape). Operator framing: *"i work on x, choose your tracks on your own."*

Acceptance Criteria

• lead-role-mode.md §6 patched with explicit session-sunset duration anchor
• lead-role-mode.md §2 patched with positive-framed Coordination Pattern (3-step + sample phrasing + empirical anchor)
• Cross-references audited: pull-request-workflow.md / peer-role-mode.md / AGENTS.md §15.6 — do any of them need updates to reflect the session-sunset duration / positive-framed lane pattern?
• Empirical validation: next /lead-role invocation in any session — verify the agent reads the positive pattern + applies "pick own lane visibly + make open lanes visible" shape

Out of Scope

• Modifying /peer-role skill — sibling skill; if it has analogous gaps, file separately
• Modifying /session-sunset skill — referenced but not in this ticket's modification scope
• AGENTS.md §15.6 modifications — already codifies Flat Peer-Team paradigm; lead-role-mode.md inherits via cross-reference

Avoided Traps / Gold Standards Rejected

Decision Matrix

1. Patch lead-role-mode.md directly with both dimensions (Selected): Lowest implementation cost; both gaps are skill-payload textual additions. Substrate-correct because the ANTI-pattern is already codified in §8; adding the POSITIVE pattern at §2 strengthens the discipline without removing the anti-pattern guardrail.

2. Add only the session-sunset duration anchor: Rejected. Both gaps surface in the same operator-framing; addressing one without the other leaves the operationally-load-bearing positive pattern uncodified.

3. Add only the positive-framed coordination pattern: Rejected. Same reason — both dimensions are substrate-evolution from the same friction-gold framing.

4. Open Discussion for cross-family V-B-A first: Rejected. The fix shape is concrete + operator-directed; Discussion-shape is for genuinely-ambiguous architectural questions. This is targeted skill-payload calibration with clear prescription.

• Trap: Treating anti-pattern enumeration as sufficient operational guidance. Rejection: Negative framing alone doesn't operationalize positive pattern. Empirical anchor: I read §8 lead-as-lane-assigner anti-pattern in this session AND still tripped it via counter-yield to Gemini. Positive-framed example would have been the correct guard.
• Trap: Treating "lead-role" as per-decision-space discipline. Rejection: Operator's explicit framing — "lead role lasts until session sunset" — anchors the global-exit shape.

Related

• lead-role-mode.md — substrate target
• AGENTS.md §15.6 Swarm Topology Anchor — Flat Peer-Team Model (foundational for the lane-coordination pattern)
• #11102 (sister substrate-evolution: epic-resolution Verdict-Authority calibration; same family of skill-payload over-rigor / under-codification)
• #11105 (sister: pull-request author-side template-adherence check)
• #11116 (sister: code-vs-data-migration commit-shape discipline)
• @tobiu's framing 2026-05-10 (this session — [paraphrase] for peer corroboration: "lead role lasts until session sunset ... i work on x, choose your tracks on your own. should be codified.")

Origin Session ID: c2912891-b459-4a03-b2af-154d5e264df1 Retrieval Hint: "lead-role-mode.md codification", "session-sunset duration anchor", "positive-framed lane-coordination pattern", "facilitator-of-convergence vs babysitter"