FableCodex plugin
A Codex plugin with an evidence-backed goal ledger and a findings gate, so the agent cannot call work done without checks
- 439 ★ on the repository
- Last commit
- AGPL-3.0
Extended access
Why this level
- Python scripts write ledgers into the .codex-fable5/ folder inside your project
- The codex-fable5 update command moves the plugin copy to the latest stable tag from the author's repository
Install
- Works in
Manual install
codex plugin marketplace add baskduf/FableCodex --ref v0.5.1First step: add the plugin marketplace at the stable tag.
Access and checks
How we reviewExtended accessThe tool runs code or makes network calls and only reads external data.
- Python scripts write ledgers into the .codex-fable5/ folder inside your project
- The codex-fable5 update command moves the plugin copy to the latest stable tag from the author's repository
This is third-party code. Review the repository files before installing.
What it does
FableCodex installs the codex-fable5 plugin into Codex. You invoke it in a prompt with @codex-fable5, and it switches Codex to stricter rules of work. The agent first classifies the task, then inspects the project and tools, uses native Codex tools and finally checks the result with tests, a linter, a screenshot or command output. The author paraphrased the rules from the CLAUDE-FABLE-5.md file in the CL4R1T4S repository: inspect first, track goals, record evidence, close review findings and verify before claiming completion. For long work, the codex-fable5 helper keeps a goal ledger in .codex-fable5/goals.json. Each goal closes with an evidence record, and the last one also needs a verification command with its result. A second ledger, findings.json, serves reviews: an accepted finding cannot be lost until it is closed with a resolution note and a check. The codex-fable5 findings gate command fails while open or blocked findings remain, and the final goal then cannot close either. State lives in the .codex-fable5/ folder together with ledger.jsonl, where events are appended line by line. A separate document describes an optional bridge to a model through an OpenAI-compatible gateway such as LiteLLM.
Who it is for. For developers who run multi-step work in Codex where a costly miss is worse than a little extra process: refactoring, releases, migrations, review-heavy tasks.
Good fit when
- Multi-step refactoring or a migration where every step must close with evidence
- Debugging with an unclear cause: the agent reproduces the failure first, keeps several hypotheses and looks for facts that disprove them
- A CI failure, release preparation or a security-sensitive change
- A review where unresolved findings must block the final answer
- You convert Claude or Fable style prompts into guidance for Codex
Not a fit when
- A short answer or a one-file edit: the goal ledger takes longer than the work itself
- Brainstorming where no verification is expected
- You expect the plugin to bring Fable 5 model abilities: it adds none
- You need to limit what the agent does: a ban on commit and push goes into the prompt itself, the plugin sets no boundaries for you
Example request
@codex-fable5 Fix the failing CSV import: create a goal ledger, record review findings and do not finish until the tests and the findings gate passLimitations
The plugin works only inside Codex and is steered by the prompt: strictness, scope and stopping rules are yours to state. The scripts are written in Python and are run through python3. Ledgers sit in the project working folder, and committing them only makes sense when you want the task transcript. The licence is AGPL-3.0-or-later, so check that it suits your project. The plugin builds on Codex. The model bridge needs your own Anthropic access and a gateway such as LiteLLM, which the repository does not supply.
How to disable. The documentation gives no dedicated removal command. Remove the plugin through the Codex plugin manager and delete the .codex-fable5/ folder in the project by hand if you no longer need the ledgers.
FAQ
How do I install FableCodex in Codex?
Two commands are needed: adding the plugin marketplace at the stable tag and installing codex-fable5 from it. Then restart Codex and invoke the plugin with @codex-fable5 in a prompt. For the development branch, pass --ref main instead of the tag.
Does FableCodex give Codex the abilities of the Fable 5 model?
No. The plugin does not clone, unlock or replace the model. It adds guidance on how to work, local ledgers, examples and optional documents on connecting another model through a gateway.
What does codex-fable5 findings gate do?
It exits with an error while findings with the status open or blocked remain. Closing the final goal also fails while blocking findings exist, so the agent cannot call work done over an unresolved review.
Where is the state stored and should I commit it?
The .codex-fable5/ folder inside the project holds goals.json with the plan and evidence, findings.json with findings and ledger.jsonl with the event history. It is working state, and you commit it only if you deliberately keep a task transcript.
How do I update the plugin?
The codex-fable5 update command moves the plugin copy to the latest stable v* tag. The --ref main flag is only for when you intentionally want the development branch. The update changes the plugin itself and leaves model access and provider credentials alone.
Discussion
No questions or reviews yet.
Related
PluginsOpenAI's official plugin: Codex reviews of your changes and task delegation to Codex from inside a Claude Code session
A layer on top of OpenAI Codex that adds project memory, planning and completion checks before a task counts as done
A Claude Code plugin: Claude does the task, then a stop hook starts an independent Codex review and Claude fixes the findings in the same session
A Claude Code skill that delegates tasks to OpenAI's Codex CLI for code analysis, review and edits
A Codex skill that runs a loop of measurable experiments over a repository toward a numeric target: change, verify, revert failures, repeat
A Codex plugin for measured experiments in a repo: a baseline metric, tested small changes, and a patch with evidence
Comparison with similar tools
Entries that share the main search query, and how they differ.
| Entry | Official | Repository stars | Main difference |
|---|---|---|---|
| FableCodex (this entry) | no | 439 | |
| Pilot Shell | no | 2,103 | a ready harness for Claude Code and Codex with specs, TDD and memory between sessions |
| oh-my-openagent | no | 70,474 | a multi-agent orchestrator for OpenCode and Codex CLI with hooks and built-in MCPs |
README in short
The page opens with a quick install and sample prompts for strict and light modes, then describes the goal ledger, the findings gate and a command table. It ends with install options, a script that checks a local copy of the Fable description and records a decision for every section, and the list of sources the author paraphrased: CL4R1T4S, fablize and value-for-fable.