API Relay Audit is a local security audit tool for third-party AI API relays and LLM proxies. This policy covers vulnerabilities in this repository, not findings about relay operators.

Until the first tagged release, the supported line is the current master branch. After tagged releases begin, the supported lines are the current master branch and the latest published release. Older snapshots may be useful for comparison, but security fixes land on master first.

If you find a vulnerability in API Relay Audit itself, use GitHub private vulnerability reporting when it is enabled for this repository.

If private vulnerability reporting is not available, open a minimal public issue only with non-sensitive reproduction details. Do not include secrets, private relay details, API keys, wallet material, or raw traffic captures in a public issue.

Please include:

• The affected file, command, or workflow.
• A minimal reproduction that does not expose real API keys.
• Whether the issue affects the standalone audit.py, the modular package, or both.
• Any evidence that a report could leak credentials, send traffic to an unintended host, or misclassify an unsafe relay as clean.

Do not include live credentials, private relay URLs, wallet seed phrases, private keys, or user traffic captures.

Use the public evidence path for redacted relay audit evidence:

• Use the Submit Audit Evidence issue template.
• Include only sanitized screenshots, redacted report artifacts, hashes, provider/profile metadata, and reproducible non-sensitive summaries.
• Do not include API keys, raw headers, full response bodies, wallet material, private relay traffic, or user data.

Use the operator response path when you operate the relay being discussed:

• Use the Operator Response issue template.
• Prove domain control as requested by the template.
• Keep remediation notes public-safe and avoid posting secrets or private user traffic.

Use private vulnerability disclosure for sensitive issues:

• API key exposure caused by this repository or its workflows.
• A bug that can redirect audit traffic to an unintended host.
• A redaction failure involving raw response bodies, request headers, wallet material, or user data.
• A reproducible issue where this tool marks an unsafe or inconclusive state as clean.

In scope:

• Credential leakage caused by this tool.
• Incorrect redaction or unsafe report handling.
• A bug that silently changes the target relay URL.
• A bug that marks an inconclusive or anomalous audit step as clean.
• Drift between the standalone and modular distributions that changes audit behavior.

Out of scope:

• A relay operator's behavior unless it demonstrates a bug in this tool.
• Requests for relay recommendations or rankings.
• Vulnerabilities in upstream AI providers.
• Results from modified forks that do not match this repository.

API Relay Audit reports evidence from local probes. A LOW result is not a certificate that a relay is safe, and an inconclusive result is not clean. Use the report as one input to a broader security review.