Skip to content
View philpaz's full-sized avatar

Block or report philpaz

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Content in all repositories owned by your account will be closed.
Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. Markdown is supported. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
philpaz/README.md

A terminal runs recusal check philip-paz. Evidence: 20 years in financial services and regulated data, builds deterministic guardrails for AI agents, guitars and motocross: all PASS. Claims "10x engineer" and "AI agents should act with no one watching": REFUSE. Verdict: ALLOW, human in the loop.

Philip Paz

AI agents now read, decide, and act on their own. I frankly did not like that, so I built a small annoyance for them.

Recusal checks the evidence before an agent's tool call runs. Allow, retry, or refuse. No model in the decision, same evidence, same verdict. A hand on the reins, and a record for when a human comes back.

PyPI CI License

pip install recusal
recusal demo

It won't change the order of magnitude of anything. It's nuggets, on purpose. Contributions are welcome, and the good first issues are a friendly place to start.

That hash in the terminal is real. Check it yourself.

The verdict above is verdict.txt, with LF line endings:

curl -s https://raw.githubusercontent.com/philpaz/philpaz/main/verdict.txt | sha256sum

It should print 9475c1b34e8b28069da5f2727701716611207c8983071ac57e31fcb513c1d883. Same evidence, same verdict.

A small robot on a dirt bike rolls up to a striped gate. The sign says REFUSE, no helmet. The robot says: but I'm autonomous!

LinkedIn

Pinned Loading

  1. recusal recusal Public

    Deterministic governance for Claude and MCP tool calls. Pin approved capabilities, detect drift, and refuse unsafe or unapproved actions before execution. No model in the decision path.

    Python 4 2