Security and Privacy Reporting¶

Capability Lab handles models and records that may describe a person's capabilities, evidence, development history, or observed activity. Privacy failures therefore belong in the project's security boundary even when they are not conventional remote-code-execution vulnerabilities.

What to report privately¶

Please use a private reporting channel for issues involving:

• credentials, tokens, keys, or secrets;
• accidental publication of person-scoped records or Pilot participant data;
• paths that allow private .local/ data to enter tracked repository content;
• provenance or identity-binding failures that could silently rewrite governed records;
• authorization or authority-boundary bypasses;
• security-sensitive dependency or CI issues;
• any vulnerability whose public disclosure would expose users before a fix is available.

Reporting¶

When the repository's GitHub private vulnerability reporting / Security Advisory flow is available, use Report a vulnerability rather than a public issue.

If private vulnerability reporting is unavailable, open a public issue containing only a request for a private contact channel. Do not include exploit details, credentials, personal data, private payloads, or other sensitive information in that issue.

A useful private report includes:

• affected revision or version;
• affected component;
• reproduction steps using synthetic data where possible;
• impact and violated boundary;
• whether any real sensitive data was exposed;
• suggested remediation, if known.

Public research issues¶

Ordinary correctness bugs, reproducibility problems, documentation errors, and non-sensitive architecture questions may be reported publicly.

Scope note¶

Capability Lab is experimental research software. Its epistemic and governance checks are not a substitute for application-level authentication, operating-system security, legal authorization, professional licensing, or human review where those are required.