Catches the mistakes coding agents repeat before they merge: a pull-request gate with no model call and no API key.
Short answer: Redact passwords, tokens, API keys, and authorization headers. A secret in a log or in code ends up in every place the log goes.
Group: Observability and logging. Status: stable. Generated from the rule definitions; do not edit by hand.
Playground: https://brevio.pro/tools/agent-guard-playground?rule=R13.1
Zero secret logging: redact passwords, tokens, API keys, and authorization headers.
A secret in a log or in code ends up in every place the log goes.
const sessionToken = 'eyJhbGciOiJIUzI1NiJ9.eyJzdWIiOiIxMjM0In0.example-signature';
const sessionToken = process.env.SESSION_TOKEN;
brevio-core scan: a scan of added lines (code_rules.scan_added_lines_for_rules).A person can silence one finding with brevio-core: ignore R13.1 -- reason on the line or the line above. Without the reason it does not count. An agent cannot silence it: the agent hook and the MCP tools never honour the comment, and a pull request that adds one is reported until a reviewer allows it. Fixing the code is still the better answer.
What the scanner already leaves alone: Only a known credential format (AWS, GitHub, Slack, Stripe live, Google API key, a private key block) or a random-looking value bound to a secret, token, password or API key name is flagged; a JWT bound to such a name always is. Test files are not checked. A placeholder (example, fake, test, xxxx, your-…) is not a key. Keys that are public by design are not flagged: a name that says public or publishable, a Stripe pk_ key, and an Algolia or DocSearch key (named so, or in the five lines below such a name).
A scan reads only the lines you add, so a line that was already in the file is not flagged.
To change what the rule flags, edit its definition in dev_rules.py or code_rules/__init__.py. That is a code change and is reviewed like one.