feat: bridge Codex v0.3.3–v0.4.1 (worklist routing, numbered closeout, delegation contract)

Bridges the Codex Game Studios v0.3.3 continuity-routing rework, v0.4.0 numbered closeout, and v0.4.1 role-agent delegation contract into the OpenCode-native port.

- Introduce ## Session Worklist + ## Phase Guard routing cache in production/session-state/active.md; /resume-from-handoff compiles it, post-work closeouts read it, /studio-next deprecated to manual reference
- Require numbered closeout (Next action: + 1. (Recommended)) on 21 completion skills + AGENTS wrap-up + session-continuity Pause Procedure
- Add central Role-Agent Delegation Authorization to AGENTS.md, coordination-rules.md, director-gates.md; extend skill-test checks
- Translate upstream validate_runtime.py CLOSEOUT enforcement into run_closeout in audit.sh
- Bump VERSION 0.2.0 -> 0.3.0; CHANGELOG + README + test-evidence

Bridged source: Cipher-85/Codex-Game-Studios v0.3.3, v0.4.0, v0.4.1 (2026-07-06).

Author: Cipher-85

.opencode/VERSION

@@ -1 +1 @@
-0.2.0
+0.3.0

.opencode/audit.sh

@@ -8,6 +8,7 @@
 #   all          Run all validators (default)
 #   agents       Validate agent files
 #   skills       Validate skill files
+#   closeout     Check closeout-routing contract on completion skills
 #   runtime      Check for stale references
 #   config       Validate opencode.json
 #   hooks        Test hook scripts
@@ -24,7 +25,7 @@ command="${1:-all}"
 while [ "$#" -gt 0 ]; do
   case "$1" in
     --root) root="$(cd "$2" && pwd -P)"; shift 2 ;;
-    all|agents|skills|runtime|config|hooks|smoke|release) command="$1"; shift ;;
+    all|agents|skills|runtime|config|hooks|smoke|release|closeout) command="$1"; shift ;;
     *) shift ;;
   esac
 done
@@ -97,6 +98,67 @@ sys.exit(0)
   printf '  %d skills checked\n' "$count"
 }
 
+run_closeout() {
+  printf '\n── Closeout Routing ────────────────────────────────────────\n'
+  local -a skills=(
+    architecture-decision architecture-review art-bible asset-spec brainstorm
+    code-review consistency-check design-system dev-story gate-check help
+    map-systems project-stage-detect quick-design smoke-check story-done
+    story-readiness team-qa ux-design
+  )
+  local -a markers=(
+    "Verdict: **COMPLETE**"
+    "Verdict: COMPLETE"
+    "**Verdict: COMPLETE**"
+    "## Recommended Next Steps"
+  )
+  local -a required=(
+    "Session Worklist"
+    "production/session-state/active.md"
+    "completed work"
+    "owed verification"
+    "numbered next-action prompt"
+    "Next action:"
+    "1. (Recommended)"
+    "(Recommended)"
+  )
+  local -a forbidden=(
+    "one recommended next action"
+    "numbered choice set"
+  )
+  local checked=0
+  for s in "${skills[@]}"; do
+    local f="$root/.opencode/skills/$s/SKILL.md"
+    [ -f "$f" ] || { fail "$s (no SKILL.md)"; continue; }
+    checked=$((checked + 1))
+    # Trigger only when the skill declares a closeout marker.
+    local has_marker=0 m
+    for m in "${markers[@]}"; do
+      if rg -q -F "$m" "$f" 2>/dev/null; then has_marker=1; break; fi
+    done
+    if [ "$has_marker" -eq 0 ]; then
+      pass "$s (no closeout marker; skipped)"
+      continue
+    fi
+    local missing=() bad=()
+    for m in "${required[@]}"; do
+      if ! rg -q -F "$m" "$f" 2>/dev/null; then missing+=("$m"); fi
+    done
+    for m in "${forbidden[@]}"; do
+      if rg -q -F "$m" "$f" 2>/dev/null; then bad+=("$m"); fi
+    done
+    if [ "${#missing[@]}" -eq 0 ] && [ "${#bad[@]}" -eq 0 ]; then
+      pass "$s (closeout routing complete)"
+    else
+      local detail=""
+      [ "${#missing[@]}" -gt 0 ] && detail+=" missing: ${missing[*]}"
+      [ "${#bad[@]}" -gt 0 ] && detail+=" forbidden-present: ${bad[*]}"
+      fail "$s (closeout contract)$detail"
+    fi
+  done
+  printf '  %d closeout skills checked\n' "$checked"
+}
+
 run_runtime() {
   printf '\n── Runtime References ──────────────────────────────────────\n'
   # Check for stale Claude references
@@ -241,19 +303,21 @@ case "$command" in
   all)
     run_agents
     run_skills
+    run_closeout
     run_runtime
     run_config
     run_hooks
     run_smoke
     ;;
-  agents)  run_agents ;;
-  skills)  run_skills ;;
-  runtime) run_runtime ;;
-  config)  run_config ;;
-  hooks)   run_hooks ;;
-  smoke)   run_smoke ;;
-  release) run_release ;;
-  *) printf 'Unknown command: %s\nAvailable: all, agents, skills, runtime, config, hooks, smoke, release\n' "$command" >&2; exit 2 ;;
+  agents)   run_agents ;;
+  skills)   run_skills ;;
+  closeout) run_closeout ;;
+  runtime)  run_runtime ;;
+  config)   run_config ;;
+  hooks)    run_hooks ;;
+  smoke)    run_smoke ;;
+  release)  run_release ;;
+  *) printf 'Unknown command: %s\nAvailable: all, agents, skills, closeout, runtime, config, hooks, smoke, release\n' "$command" >&2; exit 2 ;;
 esac
 
 printf '\n── Result: %d error(s) ──\n' "$errors"

.opencode/docs/coordination-rules.md

@@ -40,11 +40,31 @@ Spawned via `Task` within a single OpenCode session. Used by all `team-*` skills
 and orchestration skills. Subagents share the session's permission context, run
 sequentially or in parallel within the session, and return results to the parent.
 
+Explicit invocation of an OpenCode Game Studios skill authorizes only the
+subagent spawns declared by that skill's workflow for that run. This
+authorization is limited to delegation; normal approval rules still govern file
+writes, commits, pushes, branch changes, design decisions, game-feel or balance
+decisions, and any agent not named by the workflow.
+
 **When to spawn in parallel**: If two subagents' inputs are independent (neither
 needs the other's output to begin), spawn both Task calls simultaneously rather
 than waiting. Example: `/review-all-gdds` Phase 1 (consistency) and Phase 2
 (design theory) are independent — spawn both at the same time.
 
+**Review-mode check**: Before spawning any director or lead gate, resolve the
+active review mode using `.opencode/docs/director-gates.md`. `solo` skips all
+director gates, `lean` runs only PHASE-GATE director gates, and `full` runs
+declared gates normally.
+
+**Runtime fallback**: OpenCode's `Task` tool does not require per-spawn consent
+for agents already named by an invoked skill's workflow. If a surface still
+requires literal user consent before the first subagent spawn, ask one
+confirmation before spawning: "This skill declares [agents/gates]. May I spawn
+those role agents for this run?" If the user declines or delegation is
+unavailable, do not invent, summarize, or simulate specialist/director verdicts.
+Report the missing delegation as skipped or blocked, then continue only where
+the workflow allows a partial result.
+
 ### Agent Teams (experimental — opt-in)
 Multiple independent OpenCode *sessions* running simultaneously, coordinated
 via a shared task list. Each session has its own context window and token budget.
@@ -70,4 +90,5 @@ When an orchestration skill spawns multiple independent agents:
 1. Issue all independent Task calls before waiting for any result
 2. Collect all results before proceeding to dependent phases
 3. If any agent is BLOCKED, surface it immediately — do not silently skip
-4. Always produce a partial report if some agents complete and others block
+4. Never replace a blocked or skipped agent with an internal simulated verdict
+5. Always produce a partial report if some agents complete and others block

.opencode/docs/director-gates.md

@@ -28,6 +28,12 @@ the verdict using the **Verdict handling** rules below.
 Review intensity controls whether director gates run. It can be set globally
 (persists across sessions) or overridden per skill run.
 
+Explicit invocation of a skill that references this document authorizes the
+director and lead subagent spawns declared by that skill, after review-mode
+filtering, for that run only. This authorization does not grant file-write,
+commit, push, branch-change, design-decision, game-feel, balance, or undeclared
+agent permission.
+
 **Global config**: `production/review-mode.txt` — one word: `full`, `lean`, or `solo`.
 Set once during `/start`. Edit the file directly to change it at any time.
 
@@ -47,6 +53,12 @@ Examples:
 | `lean` | PHASE-GATEs only (`/gate-check`) — per-skill gates skipped | **Default** — solo devs and small teams; directors review at milestones only |
 | `solo` | No director gates anywhere | Game jams, prototypes, maximum speed |
 
+`--review full` means every gate declared by the invoked skill runs normally.
+For `/gate-check`, both `full` and `lean` run all declared PHASE-GATE directors:
+CD-PHASE-GATE, TD-PHASE-GATE, PR-PHASE-GATE, and AD-PHASE-GATE. `solo` skips
+those directors and leaves the gate verdict based on artifact and quality checks
+only.
+
 **Check pattern — apply before every gate spawn:**
 
 ```
@@ -62,6 +74,10 @@ Apply the resolved mode:
 - full → spawn as normal
 ```
 
+If the runtime still requires literal delegation consent before the first spawn,
+ask once for the agents/gates that survived review-mode filtering. If consent is
+denied, mark those gates skipped or blocked; do not simulate their verdicts.
+
 ---
 
 ## Invocation Pattern (copy into any skill)
@@ -76,6 +92,10 @@ Apply the resolved mode:
 - `lean` → **skip unless this is a PHASE-GATE** (CD-PHASE-GATE, TD-PHASE-GATE, PR-PHASE-GATE, AD-PHASE-GATE). Note: `[GATE-ID] skipped — Lean mode`
 - `full` → spawn as normal
 
+After applying review mode, spawn only the gates declared by the invoked skill.
+Do not add adjacent reviewers, and do not replace skipped or unavailable gates
+with internal simulated verdicts.
+
 ```
 # Apply mode check, then:
 Spawn `[agent-name]` via Task:
@@ -765,6 +785,10 @@ introduced, or when a tech art decision affects visual style
 When a workflow requires multiple directors at the same checkpoint (most common
 at `/gate-check`), spawn all agents simultaneously:
 
+`/gate-check` is the canonical PHASE-GATE workflow. In `lean` and `full` mode,
+all four PHASE-GATE directors below run unless the runtime blocks delegation. In
+`solo` mode, they are skipped.
+
 ```
 Spawn in parallel (issue all Task calls before waiting for any result):
 1. creative-director  → gate CD-PHASE-GATE

.opencode/docs/session-continuity.md

@@ -5,9 +5,10 @@ history. Prefer small, current, file-backed state over broad transcripts.
 
 ## File Roles
 
-- `production/session-state/active.md`: live working checkpoint. Use for current
-  task, progress checklist, decisions, files touched, open questions, and owed
-  verification.
+- `production/session-state/active.md`: live working checkpoint and current
+  session routing cache. Use for current task, progress checklist, decisions,
+  files touched, open questions, owed verification, `## Session Worklist`, and
+  `## Phase Guard`.
 - `production/session-handoff.md`: canonical resume narrative when a session has
   enough state that another session should continue from it.
 - `production/session-archive.md`: historical record only. Do not read by default
@@ -25,9 +26,17 @@ Before pausing a meaningful work unit:
 
 1. Record what changed and what remains.
 2. Record verification that passed, failed, was blocked, or was not run.
-3. Preserve exact next commands only when they are known to be useful.
-4. Keep local-only notes out of tracked docs unless they are project state.
-5. Suggest `/handoff [short-label]` when installed and the next session would
+3. Read or refresh `## Session Worklist` in
+   `production/session-state/active.md` and recommend the top valid lane.
+   The final response must include completed work, verification or owed
+   verification, and a numbered next-action prompt with exactly one
+   `(Recommended)` option. Use this numeric fallback even when there is only one
+   clear next lane:
+   `Next action:` then `1. (Recommended) [action label] - [brief reason /
+   command]`. The user can reply with `1`.
+4. Preserve exact next commands only when they are known to be useful.
+5. Keep local-only notes out of tracked docs unless they are project state.
+6. Suggest `/handoff [short-label]` when installed and the next session would
    otherwise need to reconstruct context.
 
 ## Resume Procedure
@@ -39,7 +48,8 @@ On resume:
 3. Read only the bounded files named by the handoff unless a deep audit is
    needed.
 4. Verify drift-prone claims cheaply before acting on them.
-5. Continue from the saved artifact unless there is a real inconsistency.
+5. Continue from the saved `## Session Worklist` unless there is a real
+   inconsistency.
 
 ## Context Thresholds
 

.opencode/skills/architecture-decision/SKILL.md

@@ -455,3 +455,21 @@ If there are no remaining priority ADRs and no undesigned GDD systems, offer onl
 > assessment. Running it here would invalidate the review.
 
 Update any stories that were `Status: Blocked` pending this ADR to `Status: Ready`.
+
+
+## Closeout Contract
+
+Every final response from this skill must include completed work, verification
+run or owed verification, and next-lane routing. Read or refresh the
+`## Session Worklist` in `production/session-state/active.md` when present. End
+with a numbered next-action prompt using numeric format only, even when there is
+only one valid lane:
+
+```md
+Next action:
+1. (Recommended) [action label] - [brief reason / command]
+```
+
+If multiple lanes are viable, add more numbered options and keep exactly one
+`(Recommended)` option. The user can reply with `1`. Do not end with only a
+static command list.

.opencode/skills/architecture-review/SKILL.md

@@ -664,3 +664,21 @@ If any spawned agent returns BLOCKED, errors, or fails to complete:
    with grouped options — not a separate plain-text question per file.
 6. **Non-blocking** — the verdict is advisory; the user decides whether to continue
    despite CONCERNS or even FAIL findings
+
+
+## Closeout Contract
+
+Every final response from this skill must include completed work, verification
+run or owed verification, and next-lane routing. Read or refresh the
+`## Session Worklist` in `production/session-state/active.md` when present. End
+with a numbered next-action prompt using numeric format only, even when there is
+only one valid lane:
+
+```md
+Next action:
+1. (Recommended) [action label] - [brief reason / command]
+```
+
+If multiple lanes are viable, add more numbered options and keep exactly one
+`(Recommended)` option. The user can reply with `1`. Do not end with only a
+static command list.

.opencode/skills/art-bible/SKILL.md

@@ -248,3 +248,21 @@ After the art bible is approved:
 - Run `/design-system [first-system]` to start authoring per-system GDDs
 - Run `/consistency-check` once GDDs exist to validate them against the art bible's visual rules
 - Run `/create-architecture` to produce the master architecture document
+
+
+## Closeout Contract
+
+Every final response from this skill must include completed work, verification
+run or owed verification, and next-lane routing. Read or refresh the
+`## Session Worklist` in `production/session-state/active.md` when present. End
+with a numbered next-action prompt using numeric format only, even when there is
+only one valid lane:
+
+```md
+Next action:
+1. (Recommended) [action label] - [brief reason / command]
+```
+
+If multiple lanes are viable, add more numbered options and keep exactly one
+`(Recommended)` option. The user can reply with `1`. Do not end with only a
+static command list.

.opencode/skills/asset-spec/SKILL.md

@@ -351,3 +351,21 @@ Every phase follows: **Identify → Confirm → Generate → Review → Approve
 
 - Run `/asset-spec [next-context]` to continue speccing remaining systems, levels, or characters
 - Run `/asset-audit` to validate delivered assets against the written specs and identify gaps or mismatches
+
+
+## Closeout Contract
+
+Every final response from this skill must include completed work, verification
+run or owed verification, and next-lane routing. Read or refresh the
+`## Session Worklist` in `production/session-state/active.md` when present. End
+with a numbered next-action prompt using numeric format only, even when there is
+only one valid lane:
+
+```md
+Next action:
+1. (Recommended) [action label] - [brief reason / command]
+```
+
+If multiple lanes are viable, add more numbered options and keep exactly one
+`(Recommended)` option. The user can reply with `1`. Do not end with only a
+static command list.

.opencode/skills/brainstorm/SKILL.md

@@ -358,3 +358,21 @@ After the game concept is written, follow the pre-production pipeline in order:
 5. `/create-architecture` — produce the master architecture blueprint
 6. `/architecture-review` — bootstrap TR registry and Requirements Traceability Matrix
 7. `/gate-check pre-production` — validate readiness before committing to production
+
+
+## Closeout Contract
+
+Every final response from this skill must include completed work, verification
+run or owed verification, and next-lane routing. Read or refresh the
+`## Session Worklist` in `production/session-state/active.md` when present. End
+with a numbered next-action prompt using numeric format only, even when there is
+only one valid lane:
+
+```md
+Next action:
+1. (Recommended) [action label] - [brief reason / command]
+```
+
+If multiple lanes are viable, add more numbered options and keep exactly one
+`(Recommended)` option. The user can reply with `1`. Do not end with only a
+static command list.