A small Codex skill I use for long tasks so work stays traceable and less fragile.
I am sharing this as-is because it helped me avoid common execution problems:
- losing context across many batches
- forgetting to sync rules/docs after code changes
- weak checkpoints that are hard to review or roll back
- closing findings without clean evidence
If you run multi-batch remediation, audit, or refactor work, this may help.
Ruleloop enforces a practical execution loop:
- Define one clear batch scope.
- Implement the smallest safe change.
- Validate immediately (tests/lint/typecheck/build/E2E as needed).
- Capture decision-grade evidence.
- Sync rules, backlog, and logbook in the same batch.
- Create a clean checkpoint commit.
The goal is simple: keep momentum without losing governance.
skills/
ruleloop/
SKILL.md
agents/openai.yaml
references/templates.md
LICENSE
README.md
Copy the skill into your global Codex skills folder:
mkdir -p "${HOME}/.codex/skills"
cp -R skills/ruleloop "${HOME}/.codex/skills/"Call the skill from Codex when starting a long-running task:
Use $ruleloop for this multi-batch task.
Example:
Use $ruleloop to run a full remediation pass and keep rules/logbook synced after each batch.
- UI/UX remediation with many batches
- cross-file refactors that need strict traceability
- tasks where documentation drift is likely
- autonomous execution where context can be lost
Use this in the repository "About" panel:
- Description:
Codex skill for governed multi-batch execution with logbook discipline, rules sync, evidence hygiene, and safe git checkpoints.
- Topics:
codex-skillai-agentworkflowgovernanceauditremediationrunbookdeveloper-productivity
Optional:
- Website:
- your docs page or gist (if you have one)
- Social preview:
- a simple image with text:
Ruleloop - Governed Multi-Batch Execution
- a simple image with text:
This is not a universal framework.
It is just a workflow that worked well for me in real long-running tasks.
If you improve it, feel free to open a PR.