Codex seems to know the conditions required for a task to run, but it only reveals those conditions through a question-and-answer process.

What version of the Codex App are you using (From “About Codex” dialog)?

26.825.51511

What subscription do you have?

pro

What platform is your computer?

windows11

What issue are you seeing?

Summary

I am using Codex Desktop on Windows 11 with a Pro account.

The issue is that Codex seems to know the conditions required for a task to run, but it only reveals those conditions through a question-and-answer process. It does not proactively list all required conditions and the complete fix path in one clear response.

Environment

• Product: Codex Desktop for Windows
• Version: 26.825.51511
• Release date shown in app: August 30, 2026
• OS: Windows 11
• Account: ChatGPT Pro
• Model/settings tested: GPT-5.6 Sol with medium, high, and very high reasoning

Scenario

I have a task workflow with:

• an index file: openTask.md
• physical task files under an openTask folder
• an orchestrator/sub-agent workflow that executes selected tasks

Example:

openTask.md
- TASK-1
- TASK-2

openTask/
- TASK-1.md
- TASK-2.md

I wanted Codex to execute TASK-2.

Actual behavior

Codex did not execute TASK-2.

It appears to know the conditions required for the task to run, but it provides them only in response to narrow follow-up questions.

For example:

• When I asked why TASK-2 did not run, Codex said other tasks had blocking issues.
• When I asked whether removing the rule would solve it, Codex said no, because the script still requires all tasks to be valid.
• When I asked about old task references, Codex said stale references could still cause inconsistency.

Each answer was locally relevant, but Codex never converted these known conditions into a complete actionable answer.

It did not say something like:

To make TASK-2 run, these are the required conditions:

1. The blocking rule must be changed or removed.
2. The script must not require all tasks to be valid if only one selected task should run.
3. Obsolete task files must be backed up or removed.
4. Stale task references must be removed from the task index file, such as openTask.md.
5. The task index and the physical task folder must be synchronized.

Expected behavior

Since my intent was clearly to run TASK-2, Codex should have listed all conditions required for that task to run without forcing me to discover them through repeated questions.

Expected response:

To run TASK-2, the following conditions must be true:

1. No global rule should block execution because of unrelated old tasks.
2. The execution script should validate only the selected task, or all listed tasks must be valid.
3. The task index file must not contain stale task references.
4. The physical task files must match the task index.
5. Any obsolete or broken old tasks must be backed up, removed, or fixed.

If the intended behavior is to run only TASK-2, the recommended fix is:
- update the script to validate only the selected active task,
- remove obsolete task references from openTask.md,
- back up or remove obsolete task files,
- then retry TASK-2.

Why this matters

The problem is not that Codex lacks the information. The problem is that it does not organize the information into a complete task-oriented answer.

It turns the solution into a question-and-answer puzzle instead of directly explaining the required conditions for the task to run.

What steps can reproduce the bug?

Steps to reproduce
Create a task workflow where openTask.md lists multiple tasks and the openTask folder contains the corresponding task files.
Add a rule or script condition that validates all listed tasks before allowing any selected task to run.
Make one old task invalid or inconsistent.
Ask Codex to execute a different valid task, for example TASK-2.
Codex blocks the selected task because of the unrelated old task.
Ask narrow follow-up questions such as:
“If I remove the rule, will it run?”
“If I remove the old task file, will it run?”
“If I clean the references, will it run?”
Observe that Codex explains each blocker only after being asked directly, instead of proactively listing all required conditions and the complete fix path.

What is the expected behavior?

When I ask Codex to execute TASK-2, Codex should not only say that execution is blocked.

Since the goal is clear, Codex should list all known conditions required for TASK-2 to run in one complete answer.

Expected response should be similar to:

TASK-2 cannot run because the workflow currently has global validation requirements.

To make TASK-2 executable, the following conditions must be checked:

1. The rule that blocks execution because of unrelated old tasks must be changed, removed, or intentionally kept.
2. The execution script must either validate only the selected active task, or all listed tasks must be valid.
3. The task index file, such as openTask.md, must not contain stale task references.
4. The physical task files under the openTask folder must match the task index.
5. Any obsolete, broken, or stale tasks must be fixed, backed up, removed, or excluded from validation.

If the intended behavior is to execute only TASK-2, the recommended fix is:

• update the script to validate only the selected active task,
• clean stale entries from openTask.md,
• back up or remove obsolete task files,
• synchronize the task index and task folder,
• then retry TASK-2.

Additional information

No response

The report names openTask.md and the openTask folder as the reproduction workflow; start by reproducing the TASK-2 scenario with an unrelated invalid task and the described follow-up questions. Done means Codex presents the relevant execution conditions and complete fix path in one response rather than revealing them piecemeal.