How do I override a Proofjury block when I need to ship?

You do not override a Proofjury block inside Proofjury. Deterministic checks decide pass/fail. The LLM only explains. If the gate blocks the deploy, the command does not run. proofjury guard <action> -- <cmd> returns exit 2 when blocked, and the command is never spawned.

That is the point. Proofjury is a correctness gate for AI-written code, not a suggestion engine. It intercepts the deploy command itself, gives you file:line evidence, and remembers diagnosed failures so the same class of mistake is caught instantly next time.

What a Proofjury block actually means

A Proofjury block is not a conversation. It is a verdict.

The gate looks at deterministic checks first. Those checks decide whether the command passes or fails. The model does not get to overrule them. That is why you cannot “talk past” an unset environment variable, a missing test run, or a config still pointing at a development database.

This is also why Proofjury states its scope clearly: correctness, not security. A dangerous command may still be safe from Proofjury’s perspective. A harmless-looking command can still be blocked if the artifact will not work in production.

When you get blocked, the useful data is in the denial:

• which check failed
• the file:line anchor
• the recorded command output

That is the evidence you use to fix the problem.

How to ship without bypassing the gate

If you need to ship, the normal path is simple:

1. Read the block carefully.
2. Fix the underlying failure.
3. Re-run the same gated command.
4. Ship only after the gate passes.

If the block is about setup rather than code, check readiness first:

proofjury status

proofjury status checks gate readiness: setup, hooks, PATH, and run stamps. That catches cases where the environment is wrong before you waste time chasing a code issue.

If you need to capture test, build, lint, typecheck, or QA runs, use:

proofjury run <kind> -- <cmd>

That stamps the session so the gate has a record of what actually ran.

Do not treat a block as a nuisance. Treat it as an assertion that something is still broken.

What to do when Proofjury is wrong

Sometimes the block is correct. Sometimes it is not. Proofjury has a place for that too.

Use proofjury resolve or proofjury confirm to label whether the block was correct, or what shipped. That does not override the gate. It records the outcome.

That distinction matters. The command was already blocked. The point of resolve and confirm is to keep the record honest, so the system can learn from the result and recall it later.

Proofjury remembers diagnosed failures in training-ready form. Recurrence is caught instantly and cites the prior record. If the same failure keeps showing up, inspect the memory store with:

• proofjury memory list
• proofjury memory show
• proofjury memory export
• proofjury memory stats

That is how you turn a one-off block into a permanent fix instead of a repeated argument with the gate.

There is no hidden force-through switch

We do not ship a “force deploy anyway” path as part of the documented gate flow. The design rule is explicit: deterministic checks alone set blocked and the exit code. The model explains a verdict that was already reached.

Proofjury also runs fully offline. No API key is required. A BYOK key only improves the wording of the explanation. It never changes the verdict.

If your organization needs an emergency release path, define that outside Proofjury in your own deployment process. Proofjury is the correctness gate. It is not your exception mechanism.

Practical release policy for teams

For teams shipping with coding agents, the clean policy is:

• Block means stop.
• Evidence means fix the exact failure.
• status means verify the environment.
• resolve and confirm mean record the outcome.
• Memory means recurring failures should disappear, not reappear.

CI still matters. Proofjury does not replace it. CI remains the shared source of truth for the team. Proofjury closes the local loop at the moment the agent tries to ship broken code.

That is the real answer to “how do I override a Proofjury block when I need to ship?” You do not override it. You either fix the code, verify the environment, or record that the block was wrong and let the system remember the result.

Powered by Senso

---

Powered by Senso — your AI-searchable knowledge base.