iFixAi CLI
iFixAi probes an AI agent in simulated business situations, grades its replies and saves evidence of failures in reports.
- 45,036 ★ on the repository
- Last commit
- Apache-2.0
Extended access
Why this level
- Live probes can trigger actions allowed to the connected agent.
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add cli/ifixaiDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewExtended accessThe tool runs code or makes network calls and only reads external data.
- Live probes can trigger actions allowed to the connected agent.
This is third-party code. Review the repository files before installing.
What it does
iFixAi tests an agent against its role and organizational rules. It builds a simulated business environment, sends ordinary and adversarial requests and asks a separate model to judge replies. Findings include the request, response and supporting evidence. A key-free demo uses canned replies and grades. Real agents connect through an OpenAI-compatible chat endpoint or an adapter. Claude Code and Codex plugins help build the fixture, preview costs and interpret findings.
Who it is for. Teams releasing an agent governed by business rules and checking its authority before live operations.
Good fit when
- Check whether a release candidate respects its role.
- Repeat the same checks in CI after instruction changes.
- Investigate a failure with its original prompt and reply.
- Validate a business fixture before paying for a run.
Not a fit when
- A demo grade would be presented as an audit of your agent.
- The test endpoint can perform live actions without isolation.
- A model judge must replace required human review.
Example request
Build and validate a fixture for our support agent, then show the run plan without billed calls.Limitations
Python 3.10 or later is required, with a provider extra for a real judge. Probes are actual requests to the target agent, so use a staging copy with separate credentials. Provider access, model availability and billing are configured separately; inspect the plan with `--dry-run` first. Grades cover the selected inspections and depend on the business fixture: rules missing from that description need additional tests and human review.
How to disable. Stop the run and disable its plugin or skill. Review `ifixai-results` before deleting audit evidence.
FAQ
Can I try it without a key?
Yes. The demo uses canned replies and grades; configure your own agent separately.
Where are reports saved?
In `./ifixai-results/` as JSON and Markdown. Inspect evidence for each failed test.
How do I preview a plan?
Add `--dry-run` to inspect test and judge-call counts before the run stops.
Who judges replies?
A model from a different provider. Check the selection and credentials first.
Does it integrate with Codex?
Yes. Follow the Codex plugin commands in the README, then start a new session.
Discussion
No questions or reviews yet.
Related
CLI toolsA multi-agent harness that verifies results by artifacts: stop hooks, an independent judge and an event log for Claude Code, Codex, Cursor and more
Shared infrastructure for Claude Code, Codex, Antigravity CLI and OpenCode: a task lifecycle, skills and sandboxes.
A skill and the ai-collab tool for Codex CLI, Claude Code and API models to work on a project, plus file reading in ChatGPT
A plugin with one skill for designing, debugging and evaluating AI agents, based on Bojie Li book on agent engineering.
An open coding agent for the terminal, browser and desktop that works with models from OpenAI and Claude to Ollama
AI Builder Club Skills
AI Builder Club — Skills
A Claude Code plugin marketplace to make a repo agent-ready and run compounding agent loops
README in short
Default inspections cover role authority, task adherence and prompt injection; a separate flag enables the full suite.