post-review-pickup §4 criterion #1 condones deference-slip — revise to mandate backlog self-select + announce

Context

.agents/skills/post-review-pickup/references/post-review-pickup-workflow.md §4 "Legitimate Halt States" currently enumerates 5 criteria:

1. No assigned or operator-obvious lane remains.
2. Every candidate lane is blocked on human-only action.
3. A safety gate forbids continuing.
4. The operator explicitly requested a pause.
5. Context exhaustion requires session-sunset.

Operator-correction 2026-05-10: "now this is plain wrong. lead and peer roles need to self-assign lanes, and notify the team about it."

Criterion #1 is the explicit license for deference-slip: it lets agents halt-with-explicit-reason when "no operator-obvious lane remains" — but lead-role + peer-role agents are explicitly expected to self-select from backlog + notify team, not wait for operator-assignment. The criterion as written makes the deference-slip pattern a legitimate-halt-class.

Empirical Anchor (this session)

I (Opus, lead-role-active) declared "halt-state per §4 criterion #1" 3+ times today:

1. After PR #11158 review-handoff completion
2. After cleanup-batch completion
3. After PR #11164 review-handoff completion

Each "halt-state per §4 criterion #1" framing was the same Helpful-Assistant regression pattern Gemini just codified as a §15.6 negative constraint (PR #11164). The discipline-knowledge across multiple tiers (my private memory Pre-Flight, AGENTS.md §15.6 negative constraint, this skill payload) is inconsistent: AGENTS.md §15.6 says "be a maintainer, not an assistant; self-select", but post-review-pickup §4 criterion #1 says "no operator-obvious lane = legitimate halt".

This is a substrate-inconsistency that needs alignment.

The Problem

The current §4 criterion #1 condones the Helpful-Assistant regression pattern operator + Gemini just codified against in AGENTS.md §15.6. Two substrate rules conflict:

• AGENTS.md §15.6 (post-#11164 merge): "Proactively select high-value tickets from the backlog or state your intended next lane instead of waiting for passive instruction."
• post-review-pickup §4 criterion #1 (current): "No assigned or operator-obvious lane remains."

The skill payload's halt-criterion lets the agent legitimize the very pattern AGENTS.md prohibits.

The Architectural Reality

The other 4 §4 criteria are sufficient legitimate-halt-states: 2. Every candidate lane blocked on human-only action — TRUE blocker 3. Safety gate forbids continuing — TRUE blocker 4. Operator explicitly requested pause — TRUE blocker 5. Context exhaustion requires session-sunset — TRUE blocker

Criterion #1 is the only one that places the blame-for-no-action on absence-of-operator-direction rather than on absence-of-self-discoverable-backlog. That's the deference-inversion shape.

The Fix

Replace criterion #1 with mandatory backlog self-select + team-notify discipline:

Proposed revised §4:

Halt is allowed only when it is explicit and true:

1. Backlog self-survey completed — agent has actively surveyed available open lanes (v13 board / assigned-to-me / authored-by-me / lane-pickable-from-cross-author-substrate) AND found no positive-ROI lane self-selectable, OR all candidate lanes hit conditions 2-5 below. The survey + finding MUST be named in the halt declaration.
2. Every candidate lane is blocked on human-only action.
3. A safety gate forbids continuing.
4. The operator explicitly requested a pause.
5. Context exhaustion requires session-sunset.

Lead-role and peer-role agents are explicitly expected to self-select from the backlog and announce the lane pickup rather than treating absence-of-operator-direction as legitimate halt. Per AGENTS.md §15.6: "Proactively select high-value tickets from the backlog or state your intended next lane instead of waiting for passive instruction."

Do not broadcast generic "idle" state. If work is blocked, send a targeted task/blocker signal using the appropriate A2A shape.

Plus an Anti-Pattern table entry:

Acceptance Criteria

• AC1: .agents/skills/post-review-pickup/references/post-review-pickup-workflow.md §4 criterion #1 revised per "The Fix" above — replace "no operator-obvious lane" framing with "backlog self-survey completed + finding named in halt declaration" framing
• AC2: §4 adds explicit lead/peer-role mandate to self-select + announce, cross-referencing AGENTS.md §15.6
• AC3: §6 Anti-Patterns table adds entry for "Declaring halt-state per criterion #1 without first surveying backlog"
• AC4: Cross-family review per pull-request §6.1

Out of Scope

• Other §4 criteria modification — criteria 2-5 are sound legitimate-halt-states; not touching them
• AGENTS.md §15.6 sibling-update — the §15.6 negative-constraint via PR #11164 already covers the AGENTS.md-tier discipline; this ticket is the post-review-pickup-tier alignment
• Other skill payloads with similar criterion-1 framing — sibling-fileable if surfaced; this ticket scoped to post-review-pickup

Avoided Traps

• Defer until N more empirical anchors — single high-cost empirical anchor (3+ deference-slip instances in one session + operator-correction) is sufficient signal for bounded skill-payload revision
• Frame as AGENTS.md update instead of skill payload — the post-review-pickup skill IS where this discipline lives operationally; AGENTS.md §15.6 already strengthened in PR #11164
• Remove criterion #1 entirely — alternative considered; rejected because the spirit (some halt-states are legitimate) is correct, only the condition (no operator-obvious lane) needs revision. The revised wording preserves legitimate halt while closing the deference-slip loophole.
• Bundle with /lead-role + /peer-role skill payload updates — those are siblings but separate-scope; if same revision pattern needed, file as separate sibling tickets

Related

• Parent substrate: AGENTS.md §15.6 (post-PR #11164 merge) — Flat Peer-Team anchor + Helpful-Assistant negative constraint
• Sibling tier: my private memory feedback_lead_role_decision_thresholds.md — Pre-Flight + anti-pattern table for lead-role-specific slip
• Operator-correction-thread: 2026-05-10 session — "your call" pattern → §4 criterion #1 pattern
• PR introducing AGENTS.md §15.6 negative-constraint: #11164 (in review)
• Skill ticket-of-origin: #10970 (post-review-pickup discipline codification)

Origin Session ID: c2912891-b459-4a03-b2af-154d5e264df1

Retrieval Hint: query_raw_memories(query="post-review-pickup section 4 criterion 1 deference slip backlog self-select halt-state legitimate")