chore: unify default codex model to gpt-5.4 xhigh

Author: zenus

.claude-plugin/marketplace.json

@@ -7,7 +7,7 @@
     {
       "name": "humanize",
       "source": "./",
-      "description": "Humanize - An iterative development plugin that uses a Codex CLI worker (gpt-5.3-codex) for implementation and a separate Codex CLI analyzer/reviewer (gpt-5.2, non-codex) for independent review (cross-vendor style).",
+      "description": "Humanize - An iterative development plugin that uses a Codex CLI worker (gpt-5.4) for implementation and a separate Codex CLI analyzer/reviewer (gpt-5.4, non-codex) for independent review (cross-vendor style).",
       "version": "1.11.3"
     }
   ]

.claude-plugin/plugin.json

@@ -1,6 +1,6 @@
 {
   "name": "humanize",
-  "description": "Humanize - An iterative development plugin that uses a Codex CLI worker (gpt-5.3-codex) for implementation and a separate Codex CLI analyzer/reviewer (gpt-5.2, non-codex) for independent review (cross-vendor style).",
+  "description": "Humanize - An iterative development plugin that uses a Codex CLI worker (gpt-5.4) for implementation and a separate Codex CLI analyzer/reviewer (gpt-5.4, non-codex) for independent review (cross-vendor style).",
   "version": "1.11.3",
   "author": {
     "name": "humania-org"

README.md

@@ -15,9 +15,9 @@ A Claude Code plugin that provides iterative development with Codex review. Huma
 ## Core Philosophy
 
 **Iteration over Perfection**: Instead of expecting perfect output in one shot, Humanize leverages an iterative feedback loop where:
-- Codex worker executes `coding` tasks (default: `gpt-5.3-codex:xhigh`)
-- Codex analyzer executes `analyze` tasks (default: `gpt-5.2:xhigh`, non-codex)
-- Codex reviewer independently reviews progress (default: `gpt-5.2:xhigh`, non-codex)
+- Codex worker executes `coding` tasks (default: `gpt-5.4:xhigh`)
+- Codex analyzer executes `analyze` tasks (default: `gpt-5.4:xhigh`, non-codex)
+- Codex reviewer independently reviews progress (default: `gpt-5.4:xhigh`, non-codex)
 - Issues are caught and addressed early
 - Work continues until all acceptance criteria are met
 
@@ -103,7 +103,7 @@ HUMANIZE_CODEX_BYPASS_SANDBOX=true claude --plugin-dir /path/to/humanize
 ```mermaid
 flowchart LR
     Plan["Your Plan<br/>(plan.md)"] --> Worker["Task Routing<br/>(coding->Codex Worker, analyze->Codex Analyzer)"]
-    Worker --> Reviewer["Codex Reviews<br/>Summary (gpt-5.2, non-codex)"]
+    Worker --> Reviewer["Codex Reviews<br/>Summary (gpt-5.4, non-codex)"]
     Reviewer -->|Feedback Loop| Worker
     Reviewer -->|COMPLETE| Review["Code Review<br/>(codex review)"]
     Review -->|Issues Found| Worker
@@ -112,8 +112,8 @@ flowchart LR
 
 The loop has two phases:
 1. **Implementation Phase**: Execute tasks by tag, then Codex reviews summaries until COMPLETE
-   - `coding` tag -> execute via `/humanize:codex-worker` (default: `gpt-5.3-codex:xhigh`)
-   - `analyze` tag -> execute via `/humanize:ask-codex` (default: `gpt-5.2:xhigh`, non-codex)
+   - `coding` tag -> execute via `/humanize:codex-worker` (default: `gpt-5.4:xhigh`)
+   - `analyze` tag -> execute via `/humanize:ask-codex` (default: `gpt-5.4:xhigh`, non-codex)
 2. **Review Phase**: `codex review --base <branch>` checks code quality with `[P0-9]` severity markers
 
 ### Sub-Agent Cross-Review Protocol
@@ -176,7 +176,7 @@ OPTIONS:
   --track-plan-file      Indicate plan file should be tracked in git (must be clean)
   --max <N>              Maximum iterations before auto-stop (default: 42)
   --codex-model <MODEL:EFFORT>
-                         Codex model and reasoning effort (default: gpt-5.2:xhigh)
+                         Codex model and reasoning effort (default: gpt-5.4:xhigh)
   --codex-timeout <SECONDS>
                          Timeout for each Codex review in seconds (default: 5400)
   --push-every-round     Require git push after each round (default: commits stay local)
@@ -301,7 +301,7 @@ BOT FLAGS (at least one required):
 OPTIONS:
   --max <N>              Maximum iterations before auto-stop (default: 42)
   --codex-model <MODEL:EFFORT>
-                         Codex model and reasoning effort (default: gpt-5.2:xhigh, non-codex)
+                         Codex model and reasoning effort (default: gpt-5.4:xhigh, non-codex)
   --codex-timeout <SECONDS>
                          Timeout for each Codex review in seconds (default: 900)
   -h, --help             Show help message
@@ -314,7 +314,7 @@ OPTIONS:
 
 OPTIONS:
   --codex-model <MODEL:EFFORT>
-                         Codex model and reasoning effort (default: gpt-5.2:xhigh, non-codex)
+                         Codex model and reasoning effort (default: gpt-5.4:xhigh, non-codex)
   --codex-timeout <SECONDS>
                          Timeout for the Codex query in seconds (default: 3600)
   -h, --help             Show help message

agents/bitlesson-selector.md

@@ -19,7 +19,7 @@ You will receive:
 ## Cross-Agent Review Context
 
 - This agent markdown serves as the prompt specification for BitLesson selection.
-- Runtime execution happens via `scripts/bitlesson-select.sh`, which calls Codex CLI as `codex exec -m gpt-5.2 -c model_reasoning_effort=high`.
+- Runtime execution happens via `scripts/bitlesson-select.sh`, which calls Codex CLI as `codex exec -m gpt-5.4 -c model_reasoning_effort=xhigh`.
 - Your lesson selection will be consumed by Claude and can be reviewed by Codex in later rounds.
 - Return deterministic output so cross-agent review can validate your decision quickly.
 

commands/gen-plan.md

@@ -377,8 +377,8 @@ Example: "The implementation includes core feature X with basic validation"
 ## Task Breakdown
 
 Each task MUST include exactly one routing tag generated by the planning agent:
-- `coding`: implemented by Codex worker (`/humanize:codex-worker`, default: `gpt-5.3-codex:xhigh`)
-- `analyze`: executed via Codex analyzer (`/humanize:ask-codex`, default: `gpt-5.2:xhigh`, non-codex)
+- `coding`: implemented by Codex worker (`/humanize:codex-worker`, default: `gpt-5.4:xhigh`)
+- `analyze`: executed via Codex analyzer (`/humanize:ask-codex`, default: `gpt-5.4:xhigh`, non-codex)
 
 | Task ID | Description | Target AC | Tag (`coding`/`analyze`) | Depends On |
 |---------|-------------|-----------|----------------------------|------------|

commands/start-rlcr-loop.md

@@ -16,8 +16,8 @@ Execute the setup script to initialize the loop:
 This command starts an iterative development loop where:
 
 1. You execute the implementation plan with task-tag routing
-   - `coding` tasks: execute via `/humanize:codex-worker` (default: `gpt-5.3-codex:xhigh`)
-   - `analyze` tasks: execute via `/humanize:ask-codex` (default: `gpt-5.2:xhigh`, non-codex)
+   - `coding` tasks: execute via `/humanize:codex-worker` (default: `gpt-5.4:xhigh`)
+   - `analyze` tasks: execute via `/humanize:ask-codex` (default: `gpt-5.4:xhigh`, non-codex)
 2. Write a summary of your work to the specified summary file
 3. When you try to exit, Codex reviews your summary
 4. If Codex finds issues, you receive feedback and continue

docs/install-for-codex.md

@@ -25,7 +25,7 @@ Or use the unified installer directly:
 This will:
 - Sync `humanize`, `humanize-gen-plan`, and `humanize-rlcr` into `${CODEX_HOME:-~/.codex}/skills`
 - Copy runtime dependencies into `${CODEX_HOME:-~/.codex}/skills/humanize`
-- Use RLCR defaults: `codex exec` with `gpt-5.2:xhigh`, `codex review` with `gpt-5.2:high`
+- Use RLCR defaults: `codex exec` with `gpt-5.4:xhigh`, `codex review` with `gpt-5.4:xhigh`
 
 ## Verify
 

docs/install-for-kimi.md

@@ -141,7 +141,7 @@ Look for the "Skills" section in the help output.
 |--------|-------------|---------|
 | `path/to/plan.md` | Plan file path | Required (unless --skip-impl) |
 | `--max N` | Maximum iterations | 42 |
-| `--codex-model MODEL:EFFORT` | Codex model | gpt-5.2:xhigh |
+| `--codex-model MODEL:EFFORT` | Codex model | gpt-5.4:xhigh |
 | `--codex-timeout SECONDS` | Review timeout | 5400 |
 | `--base-branch BRANCH` | Base for code review | auto-detect |
 | `--full-review-round N` | Full alignment check interval | 5 |
@@ -163,7 +163,7 @@ Ensure you have `codex` CLI installed:
 codex --version
 ```
 
-The skills will use `gpt-5.2` with `xhigh` effort level by default.
+The skills will use `gpt-5.4` with `xhigh` effort level by default.
 
 ## Uninstall
 

hooks/lib/loop-common.sh

@@ -42,30 +42,14 @@ readonly FIELD_WORKTREE_TEAMS="worktree_teams"
 readonly FIELD_WORKTREE_ROOT="worktree_root"
 readonly FIELD_DELEGATION_ENFORCEMENT="delegation_enforcement"
 
-# Default Codex configuration (single source of truth - all scripts reference this)
+# Default Codex configuration (single source of truth - all scripts reference this).
 # Both use :- so scripts can override before sourcing (e.g. PR loop sets different model/effort).
-#
-# Role split (hybrid: role-based defaults + runtime-aware worker model):
-# - Analyzer/Reviewer (DEFAULT_CODEX_MODEL): gpt-5.2 in all modes (always available)
-# - Worker/Implementer (DEFAULT_CODEX_WORKER_MODEL): runtime-aware
-#     Claude Code plugin mode -> gpt-5.3-codex (available in plugin sandbox)
-#     Skill mode (Codex/Kimi) -> gpt-5.2 (gpt-5.3-codex unavailable outside plugin)
-#
-# Detection: check BASH_SOURCE path for .claude/plugins/cache/ (CLAUDE_PLUGIN_ROOT is a
-# config-time template variable, not a runtime env var, so we cannot check it directly).
-DEFAULT_CODEX_MODEL="${DEFAULT_CODEX_MODEL:-gpt-5.2}"
+# Worker, analyzer, and reviewer now share the same non-codex default model.
+DEFAULT_CODEX_MODEL="${DEFAULT_CODEX_MODEL:-gpt-5.4}"
 DEFAULT_CODEX_EFFORT="${DEFAULT_CODEX_EFFORT:-xhigh}"
 
 # Default worker configuration (used by /humanize:codex-worker).
-if [[ -z "${DEFAULT_CODEX_WORKER_MODEL:-}" ]]; then
-    _LOOP_COMMON_SELF="${BASH_SOURCE[0]:-$0}"
-    if [[ "$_LOOP_COMMON_SELF" == */.claude/plugins/cache/* ]]; then
-        DEFAULT_CODEX_WORKER_MODEL="gpt-5.3-codex"
-    else
-        DEFAULT_CODEX_WORKER_MODEL="gpt-5.2"
-    fi
-    unset _LOOP_COMMON_SELF
-fi
+DEFAULT_CODEX_WORKER_MODEL="${DEFAULT_CODEX_WORKER_MODEL:-gpt-5.4}"
 DEFAULT_CODEX_WORKER_EFFORT="${DEFAULT_CODEX_WORKER_EFFORT:-xhigh}"
 
 # Codex review markers

hooks/loop-codex-stop-hook.sh

@@ -117,11 +117,11 @@ FULL_REVIEW_ROUND="${STATE_FULL_REVIEW_ROUND:-5}"
 REVIEW_STARTED="$STATE_REVIEW_STARTED"
 # RLCR mode split:
 # - codex exec uses state codex_model/codex_effort (from loop-common.sh defaults)
-# - codex review uses same model, with fixed effort "high"
+# - codex review uses the same model and effort for consistency
 CODEX_EXEC_MODEL="${STATE_CODEX_MODEL:-$DEFAULT_CODEX_MODEL}"
 CODEX_EXEC_EFFORT="${STATE_CODEX_EFFORT:-$DEFAULT_CODEX_EFFORT}"
 CODEX_REVIEW_MODEL="${STATE_CODEX_MODEL:-$DEFAULT_CODEX_MODEL}"
-CODEX_REVIEW_EFFORT="high"
+CODEX_REVIEW_EFFORT="${STATE_CODEX_EFFORT:-$DEFAULT_CODEX_EFFORT}"
 CODEX_TIMEOUT="${STATE_CODEX_TIMEOUT:-${CODEX_TIMEOUT:-$DEFAULT_CODEX_TIMEOUT}}"
 ASK_CODEX_QUESTION="${STATE_ASK_CODEX_QUESTION:-false}"
 AGENT_TEAMS="${STATE_AGENT_TEAMS:-false}"
@@ -1370,8 +1370,8 @@ append_task_tag_routing_note() {
 ## Task Tag Routing Reminder
 
 Follow the plan's per-task routing tags strictly:
-- `coding` task -> execute via `/humanize:codex-worker` (default: `gpt-5.3-codex:xhigh`)
-- `analyze` task -> execute via `/humanize:ask-codex` (default: `gpt-5.2:xhigh`), then integrate the result
+- `coding` task -> execute via `/humanize:codex-worker` (default: `gpt-5.4:xhigh`)
+- `analyze` task -> execute via `/humanize:ask-codex` (default: `gpt-5.4:xhigh`), then integrate the result
 - Keep Goal Tracker Active Tasks columns `Tag` and `Owner` aligned with execution
 ROUTING_EOF
 }