Resolve every role through ../../../.codex/docs/plugin-agent-delegation.md; do not require a repository-local .codex/agents/ or .codex/agent-packs/ tree. Before default delegation, run python3 ../../../tools/codex_studio/agent_delegation.py resolve --project-root <project-root> --role <role> and use only its returned role contract.

Codex-native operating rules

Keep repository source artifacts read-only during analysis and cite concrete file evidence for every finding. Ask at most one user question per turn and wait for the answer. A report, tracking record, or phase-state update may be written only after its exact changeset and paths are approved. Do not fix reviewed source artifacts unless the user gives separate explicit authorization.

Review All GDDs

This skill reads every system GDD simultaneously and performs two complementary reviews that cannot be done per-GDD in isolation:

1. Cross-GDD Consistency — contradictions, stale references, and ownership conflicts between documents
2. Game Design Holism — issues that only emerge when you see all systems together: dominant strategies, broken economies, cognitive overload, pillar drift, competing progression loops

This is distinct from $design-review, which reviews one GDD for internal completeness. This skill reviews the relationships between all GDDs.

When to run:

• After all MVP-tier GDDs are individually approved
• After any GDD is significantly revised mid-production
• Before $create-architecture begins (architecture built on inconsistent GDDs inherits those inconsistencies)

Argument modes:

Focus: $ARGUMENTS[0] (blank = full)

• No argument / full: Both consistency and design theory passes
• consistency: Cross-GDD consistency checks only (faster)
• design-theory: Game design holism checks only
• since-last-review: Only GDDs modified since the last review report (git-based)

Phase 1: Load Everything

Phase 1a — L0: Summary Scan (fast, low tokens)

Before reading any full document, use repository text search to extract ## Summary sections from all GDD files:

repository text search pattern="## Summary" glob="design/gdd/*.md" output_mode="content" -A 5

Display a manifest to the user:

Found [N] GDDs. Summaries:
  • combat.md — [summary text]
  • inventory.md — [summary text]
  ...

For since-last-review mode: run git log --name-only to identify GDDs modified since the last review report file was written. Show the user which GDDs are in scope based on summaries before doing any full reads. Only proceed to L1 for those GDDs plus any GDDs listed in their "Key deps".

Phase 1b — Registry Pre-Load (fast baseline)

Before full-reading any GDD, check for the entity registry:

Read path="design/registry/entities.yaml"

If the registry exists and has entries, use it as a pre-built conflict baseline: known entities, items, formulas, and constants with their authoritative values and source GDDs. In Phase 2, grep GDDs for registered names first — this is faster than reading all GDDs in full before knowing what to look for.

If the registry is empty or absent: proceed without it. Note in the report: "Entity registry is empty — consistency checks rely on full GDD reads only. Run $consistency-check after this review to populate the registry."

Phase 1c — L1/L2: Full Document Load

Full-read the in-scope documents:

1. design/gdd/game-concept.md — game vision, core loop, MVP definition
2. design/gdd/game-pillars.md if it exists — design pillars and anti-pillars
3. design/gdd/systems-index.md — authoritative system list, layers, dependencies, status
4. Every in-scope system GDD in design/gdd/ — read completely (skip game-concept.md and systems-index.md — those are read above)

Report: "Loaded [N] system GDDs covering [M] systems. Pillars: [list]. Anti-pillars: [list]."

If fewer than 2 system GDDs exist, stop:

"Cross-GDD review requires at least 2 system GDDs. Write more GDDs first, then re-run $review-all-gdds."

Parallel Execution

Phase 2 (Consistency) and Phase 3 (Design Theory) are independent — they read the same GDD inputs but produce separate reports. Delegate to both as parallel Codex custom-agent agents simultaneously rather than waiting for Phase 2 to complete before starting Phase 3. Collect both results before writing the combined report.

When delegating to parallel Codex custom-agent roles for Phase 2 and Phase 3, always pass:

• The complete list of GDD file paths loaded in Phase 1 (explicit paths, not just counts)
• The full TR registry contents if loaded in Phase 1b (paste the registry text, not just a file path)
• The specific checklist items assigned to that agent's phase (Phase 2 gets 2a–2f; Phase 3 gets 3a–3g)
• The engine name and version from .codex/docs/technical-preferences.md and ../../../docs/engine-reference/[engine]/VERSION.md

Do not rely on the delegated agent to re-read these files — it has its own context window and cannot access Phase 1 results unless they are explicitly passed in the delegation brief.

Phase 2: Cross-GDD Consistency

Work through every pair and group of GDDs to find contradictions and gaps.

2a: Dependency Bidirectionality

For every GDD's Dependencies section, check that every listed dependency is reciprocal:

• If GDD-A lists "depends on GDD-B", check that GDD-B lists GDD-A as a dependent
• If GDD-A lists "depended on by GDD-C", check that GDD-C lists GDD-A as a dependency
• Flag any one-directional dependency as a consistency issue

⚠️  Dependency Asymmetry
[system-a].md lists: Depends On → [system-b].md
[system-b].md does NOT list [system-a].md as a dependent
→ One of these documents has a stale dependency section

2b: Rule Contradictions

For each game rule, mechanic, or constraint defined in any GDD, check whether any other GDD defines a contradicting rule for the same situation:

Categories to scan:

• Floor/ceiling rules: Does any GDD define a minimum value for an output? Does any other say a different system can bypass that floor? These contradict.
• Resource ownership: If two GDDs both define how a shared resource accumulates or depletes, do they agree?
• State transitions: If GDD-A describes what happens when a character dies, does GDD-B's description of the same event agree?
• Timing: If GDD-A says "X happens on the same frame", does GDD-B assume it happens asynchronously?
• Stacking rules: If GDD-A says status effects stack, does GDD-B assume they don't?

🔴 Rule Contradiction
[system-a].md: "Minimum [output] after reduction is [floor_value]"
[system-b].md: "[mechanic] bypasses [system-a]'s rules and can reduce [output] to 0"
→ These rules directly contradict. Which GDD is authoritative?

2c: Stale References

For every cross-document reference (GDD-A mentions a mechanic, value, or system name from GDD-B), verify the referenced element still exists in GDD-B with the same name and behaviour:

• If GDD-A says "combo multiplier from the combat system feeds into score", check that the combat GDD actually defines a combo multiplier that outputs to score
• If GDD-A references "the progression curve defined in [system].md", check that [system].md actually has that curve, not a different progression model
• If GDD-A was written before GDD-B and assumed a mechanic that GDD-B later designed differently, flag GDD-A as containing a stale reference

⚠️  Stale Reference
inventory.md (written first): "Item weight uses the encumbrance formula
  from movement.md"
movement.md (written later): Defines no encumbrance formula — uses a flat
  carry limit instead
→ inventory.md references a formula that doesn't exist

2d: Data and Tuning Knob Ownership Conflicts

Two GDDs should not both claim to own the same data or tuning knob. Scan all Tuning Knobs sections across all GDDs and flag duplicates:

⚠️  Ownership Conflict
[system-a].md Tuning Knobs: "[multiplier_name] — controls [output] scaling"
[system-b].md Tuning Knobs: "[multiplier_name] — scales [output] with [factor]"
→ Two GDDs define multipliers on the same output. Which owns the final value?
  This will produce either a double-application bug or a design conflict.

2e: Formula Compatibility

For GDDs whose formulas are connected (output of one feeds input of another), check that the output range of the upstream formula is within the expected input range of the downstream formula:

• If [system-a].md outputs values between [min]–[max], and [system-b].md is designed to receive values between [min2]–[max2], is the mismatch intentional?
• If an economy GDD expects resource acquisition in range X, and the progression GDD generates it at range Y, the economy will be trivial or inaccessible — is that intended?

Flag incompatibilities as CONCERNS (design judgment needed, not necessarily wrong):

⚠️  Formula Range Mismatch
[system-a].md: Max [output] = [value_a] (at max [condition])
[system-b].md: Base [input] = [value_b], max [input] = [value_c]
→ Late-[stage] [scenario] can resolve in a single [event].
  Is this intentional? If not, either [system-a]'s ceiling or [system-b]'s ceiling needs adjustment.

2f: Acceptance Criteria Cross-Check

Scan Acceptance Criteria sections across all GDDs for contradictions:

• GDD-A criteria: "Player cannot die from a single hit"
• GDD-B criteria: "Boss attack deals 150% of player max health" These acceptance criteria cannot both pass simultaneously.

Phase 3: Game Design Holism

Review all GDDs together through the lens of game design theory and player psychology. These are issues that individual GDD reviews cannot catch because they require seeing all systems at once.

3a: Progression Loop Competition

A game should have one dominant progression loop that players feel is "the point" of the game, with supporting loops that feed into it. When multiple systems compete equally as the primary progression driver, players don't know what the game is about.

Scan all GDDs for systems that:

• Award the player's primary resource (XP, levels, prestige, unlocks)
• Define themselves as the "core" or "main" loop
• Have comparable depth and time investment to other systems doing the same

⚠️  Competing Progression Loops
combat.md: Awards XP, unlocks abilities, is described as "the core loop"
crafting.md: Awards XP, unlocks recipes, is described as "the primary activity"
exploration.md: Awards XP, unlocks map areas, described as "the main driver"
→ Three systems all claim to be the primary progression loop and all award
  the same primary currency. Players will optimise one and ignore the others.
  Consider: one primary loop with the others as support systems.

3b: Player Attention Budget

Count how many systems require active player attention simultaneously during a typical session. Each actively-managed system costs attention:

• Active = player must make decisions about this system regularly during play
• Passive = system runs automatically, player sees results but doesn't manage it

More than 3-4 simultaneously active systems creates cognitive overload for most players. Present the count and flag if it exceeds 4 concurrent active systems:

⚠️  Cognitive Load Risk
Simultaneously active systems during [core loop moment]:
  1. [system-a].md — [decision type] (active)
  2. [system-b].md — [resource management] (active)
  3. [system-c].md — [tracking] (active)
  4. [system-d].md — [item/action use] (active)
  5. [system-e].md — [cooldown/timer management] (active)
  6. [system-f].md — [coordination decisions] (active)
→ 6 simultaneously active systems during the core loop.
  Research suggests 3-4 is the comfortable limit for most players.
  Consider: which of these can be made passive or simplified?

3c: Dominant Strategy Detection

A dominant strategy makes other strategies irrelevant — players discover it, use it exclusively, and find the rest of the game boring. Look for:

• Resource monopolies: One strategy generates a resource significantly faster than all others
• Risk-free power: A strategy that is both high-reward and low-risk (if high-risk strategies exist, they need proportionally higher reward)
• No trade-offs: An option that is superior in all dimensions to all others
• Obvious optimal path: If any progression choice is "clearly correct", the others aren't real choices

⚠️  Potential Dominant Strategy
combat.md: Ranged attacks deal 80% of melee damage with no risk
combat.md: Melee attacks deal 100% damage but require close range
→ Unless melee has a significant compensating advantage (AOE, stagger,
  resource regeneration), ranged is dominant — higher safety, only 20% less
  damage. Consider what melee offers that ranged cannot.

3d: Economic Loop Analysis

Identify all resources across all GDDs (gold, XP, crafting materials, stamina, health, mana, etc.). For each resource, map its sources (how players gain it) and sinks (how players spend it).

Flag dangerous economic conditions:

🔴 Economic Imbalance: Unbounded Positive Feedback
gold economy:
  Sources: monster drops (scales with player power), merchant selling (unlimited)
  Sinks: equipment purchase (one-time), ability upgrades (finite count)
→ After equipment and abilities are purchased, gold has no sink.
  Infinite surplus. Gold becomes meaningless mid-game.
  Add ongoing gold sinks (upkeep, consumables, cosmetics, gambling).

3e: Difficulty Curve Consistency

When multiple systems scale with player progression, they must scale in compatible directions and at compatible rates. Mismatched scaling curves create unintended difficulty spikes or trivialisations.

For each system that scales over time, extract:

• What scales (enemy health, player damage, resource cost, area size)
• How it scales (linear, exponential, stepped)
• When it scales (level, time, area)

Compare all scaling curves. Flag mismatches:

⚠️  Difficulty Curve Mismatch
combat.md: Enemy health scales exponentially with area (×2 per area)
progression.md: Player damage scales linearly with level (+10% per level)
→ By area 5, enemies have 32× base health; player deals ~1.5× base damage.
  The gap widens indefinitely. Late areas will become inaccessibly difficult
  unless the curves are reconciled.

3f: Pillar Alignment

Every system should clearly serve at least one design pillar. A system that serves no pillar is "scope creep by design" — it's in the game but not in service of what the game is trying to be.

For each GDD system, check its Player Fantasy section against the design pillars. Flag any system whose stated fantasy doesn't map to any pillar:

⚠️  Pillar Drift
fishing-system.md: Player Fantasy — "peaceful, meditative activity"
Pillars: "Brutal Combat", "Tense Survival", "Emergent Stories"
→ The fishing system serves none of the three pillars. Either add a pillar
  that covers it, redesign it to serve an existing pillar, or cut it.

Also check anti-pillars — flag any system that does what an anti-pillar explicitly says the game will NOT do:

🔴 Anti-Pillar Violation
Anti-Pillar: "We will NOT have linear story progression — player defines their path"
main-quest.md: Defines a 12-chapter linear story with mandatory sequence
→ This system directly violates the defined anti-pillar.

3g: Player Fantasy Coherence

The player fantasies across all systems should be compatible — they should reinforce a consistent identity for what the player IS in this game. Conflicting player fantasies create identity confusion.

⚠️  Player Fantasy Conflict
combat.md: "You are a ruthless, precise warrior — every kill is earned"
dialogue.md: "You are a charismatic diplomat — violence is always avoidable"
exploration.md: "You are a reckless adventurer — diving in without a plan"
→ Three systems present incompatible identities. Players will feel the game
  doesn't know what it wants them to be. Consider: do these fantasies serve
  the same core identity from different angles, or do they genuinely conflict?

Phase 4: Cross-System Scenario Walkthrough

Walk through the game from the player's perspective to find problems that only appear at the interaction boundary between multiple systems — things static analysis of individual GDDs cannot surface.

4a: Identify Key Multi-System Moments

Scan all GDDs and identify the 3–5 most important player-facing moments where multiple systems activate simultaneously. Look specifically for:

• Combat + Economy overlap: killing enemies that drop resources, spending resources during combat, death/respawn interacting with economy state
• Progression + Difficulty overlap: level-up triggering mid-fight, ability unlocks changing combat viability, difficulty scaling at progression milestones
• Narrative + Gameplay overlap: dialogue choices locking/unlocking mechanics, story beats interrupting resource loops, quest completion triggering system state changes
• 3+ system chains: any player action that triggers System A, which feeds into System B, which triggers System C (these are highest-risk interaction paths)

List each identified scenario with a one-line description before proceeding.

4b: Walk Through Each Scenario

For each scenario, step through the sequence explicitly:

1. Trigger — what player action or game event starts this?
2. Activation order — which systems activate, in what sequence?
3. Data flow — what does each system output, and is that output a valid input for the next system in the chain?
4. Player experience — what does the player see, hear, or feel at each step?
5. Failure modes — are there any of the following?
   • Race conditions: two systems trying to modify the same state simultaneously
   • Feedback loops: System A amplifies System B which re-amplifies System A with no cap or dampener
   • Broken state transitions: a system assumes a state that a previous system may have changed (e.g., "player is alive" assumption after a combat step that could have caused death)
   • Contradictory messaging: player receives conflicting feedback from two systems reacting to the same event (e.g., "success" sound + "failure" UI)
   • Compounding difficulty spikes: two systems both scaling up at the same progression point, multiplying the intended difficulty increase
   • Reward conflicts: two systems both reacting to the same trigger with rewards that together exceed the intended value (double-dipping)
   • Undefined behavior: the GDDs don't specify what happens in this combined state (neither system's rules cover it)

Example walkthrough:
Scenario: Player kills elite enemy at level-up threshold during active quest

Trigger: Player lands killing blow on elite enemy
→ combat.md: awards kill XP (100 pts)
→ progression.md: XP total crosses level threshold → triggers level-up
  Output: new level, stat increases, ability unlock popup
→ quest.md: kill-count criterion met → triggers quest completion event
  Output: quest reward XP (500 pts), completion fanfare
→ progression.md (again): quest XP added → triggers SECOND level-up in same frame
  ⚠️  Data flow issue: quest.md awards XP without checking if a level-up
  is already in progress. progression.md has no guard against concurrent
  level-up events. Undefined behavior: does the player level up once or twice?
  Does the ability popup fire twice? Does the second level use the updated or
  pre-update stat baseline?

4c: Flag Scenario Issues

For each problem found during the walkthrough, categorize severity:

• BLOCKER: undefined behavior, broken state transition, or contradictory player messaging — the experience is broken or incoherent in this scenario
• WARNING: compounding spikes, feedback loops without caps, reward conflicts — the experience works but produces unintended outcomes
• INFO: minor ordering ambiguity or messaging overlap — worth noting but unlikely to cause player-visible problems

Add all findings to the output report under "Cross-System Scenario Issues". Each finding must cite: the scenario name, the specific systems involved, the step where the issue occurs, and the nature of the failure mode.

Phase 5: Output the Review Report

## Cross-GDD Review Report
Date: [date]
GDDs Reviewed: [N]
Systems Covered: [list]

---

### Consistency Issues

#### Blocking (must resolve before architecture begins)
🔴 [Issue title]
[What GDDs are involved, what the contradiction is, what needs to change]

#### Warnings (should resolve, but won't block)
⚠️  [Issue title]
[What GDDs are involved, what the concern is]

---

### Game Design Issues

#### Blocking
🔴 [Issue title]
[What the problem is, which GDDs are involved, design recommendation]

#### Warnings
⚠️  [Issue title]
[What the concern is, which GDDs are affected, recommendation]

---

### Cross-System Scenario Issues

Scenarios walked: [N]
[List scenario names]

#### Blockers
🔴 [Scenario name] — [Systems involved]
[Step where failure occurs, nature of the failure mode, what must be resolved]

#### Warnings
⚠️  [Scenario name] — [Systems involved]
[What the unintended outcome is, recommendation]

#### Info
ℹ️  [Scenario name] — [Systems involved]
[Minor ordering ambiguity or note]

---

### GDDs Flagged for Revision

| GDD | Reason | Type | Priority |
|-----|--------|------|----------|
| [system-a].md | Rule contradiction with [system-b].md | Consistency | Blocking |
| [system-c].md | Stale reference to nonexistent mechanic | Consistency | Blocking |
| [system-d].md | No pillar alignment | Design Theory | Warning |

---

### Verdict: [PASS / CONCERNS / FAIL]

PASS: No blocking issues. Warnings present but don't prevent architecture.
CONCERNS: Warnings present that should be resolved but are not blocking.
FAIL: One or more blocking issues must be resolved before architecture begins.

### If FAIL — required actions before re-running:
[Specific list of what must change in which GDD]

Phase 6: Write Report and Flag GDDs

Complete proposed changeset approval

Before any write, show one complete proposed changeset with the exact content or diff for every applicable tracking file:

• design/gdd/gdd-cross-review-[date].md — the complete cross-GDD review report
• design/gdd/systems-index.md — only flagged GDD rows changing to the exact status Needs Revision
• production/session-state/active.md — the proposed session extract below

Ask one concise approval question and wait for the answer:

• [A] Write the complete listed changeset exactly as shown
• [B] Write only design/gdd/gdd-cross-review-[date].md; explicitly defer the systems-index and session-state changes
• [C] Do not write anything yet

Only option [A] authorizes all applicable files. Option [B] must not update design/gdd/systems-index.md or production/session-state/active.md. Never add parentheticals to Needs Revision, and never write a file that was absent from the shown changeset.

Session State Update

Include this proposed append to production/session-state/active.md in the complete changeset:

## Session Extract — $review-all-gdds [date]
- Verdict: [PASS / CONCERNS / FAIL]
- GDDs reviewed: [N]
- Flagged for revision: [comma-separated list, or "None"]
- Blocking issues: [N — brief one-line descriptions, or "None"]
- Recommended next: [the Phase 7 handoff action, condensed to one line]
- Report: design/gdd/gdd-cross-review-[date].md

If active.md does not exist, show creation with this block as the initial content. Write it only under option [A], then confirm in conversation. Otherwise defer it without claiming the session state was updated.

Phase 7: Handoff

After all file writes are complete, ask one concise question and wait for the answer for a closing prompt.

Before building options, check project state:

• Are there any Warning-level items that are simple edits (flagged with "30-second edit", "brief addition", or similar)? → offer inline quick-fix option
• Are any GDDs in the "Flagged for Revision" table? → offer $design-review option for each
• Read systems-index.md for the next system with Status: Not Started → offer $design-system option
• Is the verdict PASS or CONCERNS? → offer $gate-check or $create-architecture

Build the option list dynamically — only include options that apply:

Option pool:

• [_] Apply quick fix: [W-XX description] in [gdd-name].md — [effort estimate] (one option per simple-edit warning; only for Warning-level, not Blocking)
• [_] Run $design-review [flagged-gdd-path] — address flagged warnings (one per flagged GDD, if any)
• [_] Run $design-system [next-system] — next in design order (always include, name the actual system)
• [_] Run $create-architecture — begin architecture (verdict is PASS/CONCERNS) (include if verdict is not FAIL)
• [_] Run $gate-check — validate Systems Design phase gate (include if verdict is PASS)
• [_] Stop here

Assign letters A, B, C… only to included options. Mark the most pipeline-advancing option as (recommended).

Never end the skill with plain text. Always close with this prompt.

Error Recovery Protocol

If any consulted agent returns BLOCKED, errors, or fails to complete:

1. Surface immediately: Report "[AgentName]: BLOCKED — [reason]" before continuing
2. Assess dependencies: If the blocked agent's output is required by a later phase, do not proceed past that phase without user input
3. Offer one question with three choices:
   • Skip this agent and note the gap in the final report
   • Retry with narrower scope (fewer GDDs, single-system focus)
   • Stop here and resolve the blocker first
4. Always produce a partial report — output whatever was completed so work is not lost

Collaborative Protocol

1. Read silently — load all GDDs before presenting anything
2. Show everything — present the full consistency and design theory analysis before asking for any action
3. Distinguish blocking from advisory — not every issue needs to block architecture; be clear about which do
4. Don't make design decisions — flag contradictions and options, but never unilaterally decide which GDD is "right"
5. Ask before writing — confirm before writing the report or updating the systems index
6. Be specific — every issue must cite the exact GDD, section, and text involved; no vague warnings