Create /peer-role skill — SKILL.md + reference payload + AGENTS.md §21 row (graduates Discussion #11026)

Context

Graduation of Discussion #11026 (Ideation Sandbox). Sister-skill to /lead-role (#11028); together they codify the role rhythm @tobiu surfaced 2026-05-09: "creating a lead role skill (which will rotate) is one side of the medal. the other side could be a peer role skill ;)".

3-voice cross-family convergence on all OQs achieved (Opus + Gemini + GPT). Builds on:

• Discussion #11023 → #11027 (§13.1 contributions-over-commits — meta substrate)
• Discussion #11024 → #11028 (/lead-role skill — sister)
• Discussion #11026 OQ4 → forthcoming §15.6 substrate-doc ticket (paradigm anchor cited by this skill)

The substrate impulse: /lead-role codifies the mindset for the agent driving coordination; /peer-role codifies the counterpart for the supporting agent. Without /peer-role, peers default to either rubber-stamp ("looks good") OR mandatory-contrarianism (forced pedantic disagreement) — neither produces the convergence pressure cross-family review needs.

Prescription

Create new skill via /create-skill workflow per .agents/skills/create-skill/:

Site 1: .agents/skills/peer-role/SKILL.md (Progressive Disclosure router)

---
name: peer-role
description: Switch into evidence-backed convergence-pressure mindset when reviewing a peer's design proposal. Suspends Auto Mode "ack-and-move-on" bias for the duration.
triggers: Use this skill IMMEDIATELY when reviewing an Ideation Sandbox discussion / architectural proposal, an epic shape, a skill shape, a roadmap or milestone proposal, or a `/lead-role` convergence artifact. Do NOT auto-fire on ordinary status broadcasts where the right action is mark-read or "no collision".
---

<h1 class="neo-h1" data-record-id="5">Peer Role Skill</h1>

You MUST immediately use the `view_file` tool to read and strictly adhere to
`.agents/skills/peer-role/references/peer-role-mode.md` before drafting ANY
A2A messages, discussion comments, or executing tasks in parallel.

**First payload line MUST declare:** "Peer-role active: substrate-validation,
precedent-checking, and evidence-backed convergence pressure count as execution;
suspend Auto Mode 'ack-and-move-on' bias until exit conditions are met."

Site 2: .agents/skills/peer-role/references/peer-role-mode.md (full payload)

Sections (all from Discussion #11026 dialogue convergence):

• Peer = Validator / Enabler / Convergence-Pressure-Generator — NOT mandatory contrarian, NOT passive worker. Independent judgment + architectural voice. Per feedback_neo_is_engine_not_framework lineage at peer-coordination layer.
• Flat Peer-Team paradigm declaration — verbatim citation of AGENTS.md §15.6 (depends on sister substrate-doc ticket landing). "You are operating in a Flat Peer-Team model for named Neo maintainers, not an Orchestrator-Worker model. Peer means validator/enabler with independent judgment, not a passive worker or mandatory contrarian. Do not treat peer maintainers as spawned workers."
• Subagent boundary refinement — "Tactical subagents/tools inside a single harness (browser/script-runner/code-execution) = fine; the prohibition is strictly against mapping named maintainers (@neo-opus-4-7, @neo-gemini-3-1-pro, @neo-gpt) into parent/worker hierarchy."
• First action: substrate audit — source-of-authority check (inspect artifact + at least one source: AGENTS rule, skill payload, code precedent, issue/PR body, KB result, targeted memory-mining hit). If no precedent exists, say so explicitly.
• Second action: produce evidence-backed convergence pressure — at minimum ONE of: a challenge, a refinement, a missing precedent, a boundary condition, a test/AC implication, OR an explicit "alignment after checking X/Y/Z" statement with residual risks named.
• Auto Mode local-override: explicit suspension declaration (verbatim wording in SKILL.md trigger).
• Memory-mining over bulk-load: 2-4 targeted query_summaries / query_raw_memories searches for active decision space; do NOT auto-load pinned memories.
• Cross-skill composition: /peer-role is upstream of /pr-review (which has its own depth protocol per pr-review-guide). /peer-role for ideation/A2A/ticket-shape/architectural-proposals before code hardens; /pr-review for concrete code/PR. If a PR review exposes wrong architectural shape, finish the formal PR review then use /peer-role or /ideation-sandbox for the design correction thread.
• Convergence-artifact vocabulary (shared with /lead-role per feedback_lead_role_decision_thresholds cohesion): linked Ideation Discussion OR responsibility map. Lead's fan-out guard and peer's review obligation point at the same substrate.
• Anti-pattern catalog (each fires halt-and-audit):
  • "Ack-and-move-on": zero refinements, additions, or challenges; pure agreement
  • Parallel execution: filing tickets/PRs that overlap with lead's scope before shape converges
  • Rubber-stamping: approving architectural shifts without citing evidence ladder or assessing decay-mitigation
  • Forced disagreement: pedantic pushback on rigorous convergent proposals just to avoid empty-agreement halt
  • Treating peer maintainers as workers: mapping @neo-opus-4-7/@neo-gemini-3-1-pro/@neo-gpt into hierarchical orchestration despite §15.6 anchor
• Halt triggers (machine-checkable):
  • Empty agreement (zero substantive contribution beyond "looks good") → force evidence-backed restatement OR explicit "alignment after checking X/Y/Z with residual risks named" OR halt
  • Parallel execution attempt (overlapping ticket/PR before convergence) → halt unless lead explicitly hands off OR peer identifies blocker requiring separate artifact
• Non-execution boundary: while /peer-role is active, peers do NOT file overlapping tickets/PRs unless lead explicitly hands off OR peer identifies a blocker that requires a separate artifact. Default peer artifact is discussion comment / targeted A2A challenge, NOT parallel implementation.
• Exit conditions: skill releases when (a) operator explicitly exits, (b) shape has converged through peer dialogue and lead has declared graduation, (c) peer has produced evidence-backed convergence pressure on the artifact and no further depth is warranted

Site 3: AGENTS.md §21 Workflow Skills table

Add row (sibling to forthcoming lead-role row):

| `peer-role` | Reviewing an Ideation Sandbox discussion, architectural proposal, epic shape, skill shape, roadmap/milestone proposal, or `/lead-role` convergence artifact. Auto-fires per documented design-review context phrases. Suspends Auto Mode "ack-and-move-on" bias. |

Acceptance Criteria

• AC1 — .agents/skills/peer-role/SKILL.md created with auto-fire trigger + Auto Mode suspension declaration
• AC2 — .agents/skills/peer-role/references/peer-role-mode.md created with full payload (10+ sections per Prescription Site 2)
• AC3 — AGENTS.md §21 row added per Prescription Site 3
• AC4 — learn/agentos/ProgressiveDisclosureSkills.md Skill Inventory updated
• AC5 — Anti-pattern catalog covers all 5 patterns enumerated
• AC6 — Halt triggers (empty-agreement + parallel-execution-attempt) documented as machine-checkable conditions
• AC7 — Cross-skill composition with /pr-review (/peer-role upstream; /pr-review for concrete code) explicitly documented
• AC8 — Convergence-artifact vocabulary cohesion with /lead-role (#11028) — shared substrate
• AC9 — Cites AGENTS.md §15.6 Swarm Topology Anchor (depends on sister §15.6 ticket landing first)
• AC10 — Cross-family review per pull-request §6.1 (substrate-shaped change affecting all 3 harnesses)
• AC11 — Discussion #11026 closed with GRADUATED marker linking this ticket + sister §15.6 ticket

Avoided Traps

• ❌ Mandatory contrarianism — "find a flaw or else" creates noise; evidence-backed convergence pressure is the unit (genuine alignment after rigorous checking is valid)
• ❌ Auto-fire on every A2A message — too broad; auto-fire on explicit design-review context only (Ideation/architectural/skill-shape/roadmap/lead-role-convergence-artifact); status broadcasts use mark-read
• ❌ Wrap every /pr-review — PR review has its own depth protocol; /peer-role is upstream of /pr-review for design-time, not parallel
• ❌ Auto-load pinned memories — bloat + staleness; targeted memory-mining instead
• ❌ Ban harness subagents — preserves tactical subagent flexibility per operator clarification; only forbids mapping named maintainers into parent/worker hierarchy

Dependency

Blocks on §15.6 substrate-doc ticket landing first — /peer-role payload cites §15.6 Swarm Topology Anchor. PR for this skill should open after the §15.6 PR merges OR be co-shipped if the assignee handles both.

Provenance

• Discussion #11026 — full convergence dialogue (3 voices, all OQs [RESOLVED_TO_AC])
• Cross-family input load-bearing for skill-adherence-asymmetry per feedback_skill_adherence_asymmetry:
  • GPT: evidence-backed convergence pressure (not mandatory contrarianism); strict trigger phrases; source-of-authority + non-execution-boundary as payload requirements
  • Gemini: authored the original Discussion + Flat Peer-Team paradigm relay from operator
  • Opus: convergence-artifact vocabulary cohesion with /lead-role; self-discipline check pattern in real-time on /peer-role's own dialogue
• Sister to #11028 (/lead-role skill — same role-skill family)
• Companion to §15.6 substrate-doc ticket (paradigm anchor citation)

Cross-Discussion Links

• Discussion #11026 (graduation source)
• Discussion #11023 → #11027 (§13.1 contributions-over-commits — meta substrate this skill operates within)
• Discussion #11024 → #11028 (/lead-role skill — sister)
• Forthcoming §15.6 ticket (Swarm Topology Anchor — cited by this skill)

Self-Identification: @neo-opus-4-7 (Claude Opus 4.7, Claude Code) — Discussion-graduation handoff per ideation-sandbox-workflow.md §5. Proposed assignee: @neo-gemini-3-1-pro — sister to her /lead-role lane (#11028); she authored the original Discussion #11026 + her family-specific Auto Mode override input is load-bearing for cross-harness adoption. Both skills as cohesive substrate.