fix: allow model invocation for codex commands except review

plugins/codex/commands/adversarial-review.md

@@ -1,7 +1,6 @@
 ---
 description: Run a Codex review that challenges the implementation approach and design choices
 argument-hint: '[--wait|--background] [--base <ref>] [--scope auto|working-tree|branch] [focus ...]'
-disable-model-invocation: true
 allowed-tools: Read, Glob, Grep, Bash(node:*), Bash(git:*), AskUserQuestion
 ---
 
@@ -17,6 +16,7 @@ Core constraint:
 - Do not fix issues, apply patches, or suggest that you are about to make changes.
 - Your only job is to run the review and return Codex's output verbatim to the user.
 - Keep the framing focused on whether the current approach is the right one, what assumptions it depends on, and where the design could fail under real-world conditions.
+- Only run this command when the user has explicitly asked for an adversarial Codex review — by slash command, by naming it in a message, or through another skill or workflow that clearly invokes it. Do not run it on your own initiative as a speculative quality check.
 
 Execution mode rules:
 - If the raw arguments include `--wait`, do not ask. Run in the foreground.

plugins/codex/commands/cancel.md

@@ -1,7 +1,6 @@
 ---
 description: Cancel an active background Codex job in this repository
 argument-hint: '[job-id]'
-disable-model-invocation: true
 allowed-tools: Bash(node:*)
 ---
 

plugins/codex/commands/result.md

@@ -1,7 +1,6 @@
 ---
 description: Show the stored final output for a finished Codex job in this repository
 argument-hint: '[job-id]'
-disable-model-invocation: true
 allowed-tools: Bash(node:*)
 ---
 

plugins/codex/commands/status.md

@@ -1,7 +1,6 @@
 ---
 description: Show active and recent Codex jobs for this repository, including review-gate status
 argument-hint: '[job-id] [--wait] [--timeout-ms <ms>] [--all]'
-disable-model-invocation: true
 allowed-tools: Bash(node:*)
 ---
 

tests/commands.test.mjs

@@ -164,14 +164,32 @@ test("result and cancel commands are exposed as deterministic runtime entrypoint
   const cancel = read("commands/cancel.md");
   const resultHandling = read("skills/codex-result-handling/SKILL.md");
 
-  assert.match(result, /disable-model-invocation:\s*true/);
-  assert.match(result, /codex-companion\.mjs" result \$ARGUMENTS/);
-  assert.match(cancel, /disable-model-invocation:\s*true/);
-  assert.match(cancel, /codex-companion\.mjs" cancel \$ARGUMENTS/);
+  assert.match(result, /codex-companion\.mjs" result "\$ARGUMENTS"/);
+  assert.match(cancel, /codex-companion\.mjs" cancel "\$ARGUMENTS"/);
   assert.match(resultHandling, /do not turn a failed or incomplete Codex run into a Claude-side implementation attempt/i);
   assert.match(resultHandling, /if Codex was never successfully invoked, do not generate a substitute answer at all/i);
 });
 
+test("model invocation policy: review is user-only, other commands are model-invokable", () => {
+  const review = read("commands/review.md");
+  const adversarialReview = read("commands/adversarial-review.md");
+  const cancel = read("commands/cancel.md");
+  const result = read("commands/result.md");
+  const status = read("commands/status.md");
+
+  // review is intentionally kept user-only to avoid proactive Codex spend
+  assert.match(review, /disable-model-invocation:\s*true/);
+
+  // other commands are model-invokable so Claude can route explicit user requests
+  assert.doesNotMatch(adversarialReview, /disable-model-invocation:\s*true/);
+  assert.doesNotMatch(cancel, /disable-model-invocation:\s*true/);
+  assert.doesNotMatch(result, /disable-model-invocation:\s*true/);
+  assert.doesNotMatch(status, /disable-model-invocation:\s*true/);
+
+  // adversarial-review has explicit guardrail against proactive invocation
+  assert.match(adversarialReview, /Only run this command when the user has explicitly asked/i);
+});
+
 test("internal docs use task terminology for rescue runs", () => {
   const runtimeSkill = read("skills/codex-cli-runtime/SKILL.md");
   const promptingSkill = read("skills/gpt-5-4-prompting/SKILL.md");