Files
aetherbound-guild/.agent-taskgraph/queue/failed/ABG-V2-SYSTEMS-REVIEW-R3/goal.md
T

4.1 KiB
Raw Blame History

Goal: ABG-V2 Systems Integrity Review R3

Stage: failed fixed-revision pre-development re-review Started: 2026-08-11T08:52:00+08:00 Task ID: ABG-V2-SYSTEMS-REVIEW-R3 Baseline: d454120074057db0056bb8f117371f26de155e89; branch=review/abg-v2-systems-r3; clean Frozen spec: .agent-taskgraph/spec.md revision design-spec-r3 Graph node: .agent-taskgraph/graph.yaml#v2_systems_review_r3 revision graph-r6 Independent contract: docs/reviews/PRE_DEVELOPMENT_READINESS_AUDIT.md Context manifest: context.md Context revision: 1 Legal terminal: ABG-V2-SYSTEMS-REVIEW-R3 PASS or ABG-V2-SYSTEMS-REVIEW-R3 FAIL: or ABG-V2-SYSTEMS-REVIEW-R3 INVALID:

Objective

Independently re-audit immutable revision d454120074057db0056bb8f117371f26de155e89 from zero. Prove the recruit-shop and horizontal ordered-line auto-battle loop, economy, pacing, deterministic battle outcomes, save and failure contracts are executable and mutually consistent. Do not inherit R2 partial acceptance.

Graph Input, Output And Write Scope

  • Needs: v2_outcome_reintegration
  • Consumes: the fixed revision, design-spec-r3, current product/content authorities and current validator
  • Produces: one read-only PASS/FAIL/INVALID terminal with exact evidence
  • Writes: none in the repository; reviewer-owned temporary evidence only
  • Route: PASS -> v2_review_synthesis; FAIL/INVALID -> failed; max attempt 1

Assignment

  • Role ref: role:systems-integrity-reviewer-r3
  • 角色职责: independently verify the complete mechanical design on one immutable revision
  • 角色生命周期: task-scoped
  • 连续性: fresh R3 role and fresh HAPI session; R1/R2 are append-only failure history
  • 工作区: /Users/wusumac/Documents/game/tmp/aetherbound-guild-worktrees/review-systems-r3(分支:review/abg-v2-systems-r3)
  • Runtime requested: runtime=hapi; flavor=codex; model=gpt-5.6-sol; effort=xhigh; permission=bypassPermissions; visibility=visible
  • Runtime observed: runtime=hapi; flavor=codex; model=gpt-5.6-sol; effort=xhigh; permission=yolo; visibility=visible
  • Runtime verification: VERIFIED
  • Session evidence: HAPI afcafe40-d67b-4372-a63a-6cb94c2e6127, PID 87379, verification 7ba2650f-90fa-485d-aa7b-f2db9d38678e, runtime-evidence.json
  • Dispatch bootstrap: dispatch.md
  • Dispatch message: NOT_SENT
  • Identity ACK: PENDING

Required Gates

  1. Prove the review archive is exactly the fixed revision, the canonical repository is clean/upstream, and no later source influences judgment.
  2. Run validator self-test, full and JSON twice; require cases=15, errors=0, ok=true, exact counts and byte-identical JSON.
  3. Trace the first ten minutes and every cycle boundary with at least one legal player action in every state.
  4. Re-audit dismissal, zero-Recruit prevention, casualty replacement, explicit abandonment, stale/duplicate/interrupted operations and no-mutation failures.
  5. Recalculate targeting, order, reach, movement, protection, critical, mitigation, Raw floor, status and every deterministic AO_* terminal outcome including mixed area and duplicate-trigger suppression.
  6. Recalculate prices, rewards, sinks, XP, levels, promotion and solvency/anti-dominance fixtures from declared formulas.
  7. Reconcile retreat, defeat, casualties, collapse and abandonment with pending/applied recovery, corruption, rollback, interruption and bounded offline rules.
  8. Verify five pacing profiles, exact decision timestamps and longest no-decision gaps against the declared playtime envelope.
  9. Mechanically check all 42 professions, 36 traits, 320 equipment, 60 artifacts, 100 ordinary/elite enemies, 16 bosses and 8 regions for executable counterfactual decisions.
  10. Confirm dependent content/art/audio uses declared deterministic outcomes and introduces no probabilistic accuracy/evasion or copied commercial expression.

Frozen

  • No repository writes, repairs, commits, pushes, media, HTML, runtime, package or release actions.
  • tools/validate_design_contracts.py and prototype/site/ are superseded and are not current evidence.
  • Human fun/comprehension, runtime, device and release remain open after PASS.