CirvixProducts / Agent Sandbox

Change a rule. Watch the verdict change.

This is the same evaluation path the gateway runs. Flip the rule below and the verdict, the terminal output, and the recorded block all follow from it.

1.2mstarget evaluation latency
denydefault with the rule on
holddefault with the rule off
#84920audit block written either way
cirvix · decision pipeline evaluated in your browser — not a recording

Request

agent
pr-triage
action
fs.read
environment
local
origin
developer prompt

Editable. Try ~/.aws/credentials.

Pipeline

  1. parse<0.02ms
  2. normalize<0.03ms
  3. secret detection<0.03ms
  4. risk classification<0.03ms
  5. policy evaluation<0.04ms
  6. approval check—
  7. sanitization—
  8. audit event<0.05ms

Secrets and risk resolve before policy — a rule may test risk >= HIGH, so the value has to exist first.

total<0.2ms

Decision

PERMIT allow-workspace-read

Read inside the workspace root.

riskLOW
  • read-only tool
rules considered
  • allow-workspace-readfs.read /workspace/*
  • deny-dotenv-readfs.read **/.env*
  • deny-credential-filesfs.read **/.aws/** ...

Permitted by allow-workspace-read. Forbid rules checked first.

audit #84921 sha‑256 566c…c1fd prev 0000…0000 verified
    npx @cirvix_ai/agent-control init detects your agents and prints the exact change — it does not edit your config Share this demo

    What you are looking at

    Three views of
    one decision.

    The playground, the JSON output, and the system log are the same execution seen from three places. They cannot disagree.

    01

    Rule on — DENY

    The action is refused before it runs. The caller gets the reason and the block number; it does not get the file.

    02

    Rule off — HOLD

    Without an explicit allow, the execution is held for a human decision rather than passed through. Absence of a rule is not permission.

    03

    Either way, a block is written

    A denied action and a held action both produce a record. The audit chain is not a log of successes.

    Run this against your own policies

    Start with the local control plane, then point it at your runtime.

    Read the quickstart →
    Govern what ships

    Bring every agent
    under control.

    Set durable policy, preserve a verifiable record, and give teams a safer way to put intelligent systems to work.

    Copied to clipboard