Skip to content

Security: daniel-sc/contextveil

Security

SECURITY.md

Security Policy

ContextVeil is a local model-context redaction primitive. Please read limitations.md before reporting: several gaps are deliberate, documented parts of the V1 boundary rather than vulnerabilities.

Reporting A Vulnerability

Report suspected vulnerabilities privately through GitHub Security Advisories:

https://github.com/daniel-sc/contextveil/security/advisories/new

Please do not open a public issue for an unfixed vulnerability.

Include:

  • affected version (contextveil --version) and platform;
  • the coding-agent harness and integration involved, if any;
  • reproduction steps and observed versus expected behavior;
  • the impact you believe the issue has.

Never include a real credential. Reproduce with a generated placeholder value such as SSCANARY_EXAMPLE_0123456789abcdef. Reports containing real credentials will be deleted and you will be asked to rotate the value and resend the report.

Response Expectations

  • Acknowledgement within 5 working days.
  • An assessment, including whether the report describes a documented limitation, within 15 working days.
  • Coordinated disclosure once a fix or documented mitigation is available.

In Scope

  • Disclosure of a resolved source value through diagnostics, logs, snapshots, configuration, error output, or intervention metadata.
  • Failure to redact an enrolled value on a covered adapter path that the support matrix presents as protected.
  • Terminal escape, path, or label injection through ContextVeil output.
  • Configuration or integration installation that corrupts unrelated user or harness configuration.
  • Execution of untrusted content by the configuration, dotenv, JSON, properties, or npmrc parsers.

Out Of Scope

Everything documented in limitations.md, in particular:

  • direct credential use or network exfiltration by a local process (LIM-001);
  • unknown, transformed, or encoded values (LIM-002);
  • content outside string values, such as object keys or binary data (LIM-003);
  • destructive replacement caused by deliberately enrolling common values (LIM-004);
  • fail-open behavior of hosts that crash, disable, time out, or bypass a hook (LIM-012);
  • memory-residency of resolved values in ordinary process memory (LIM-020).

Supported Versions

The latest release in the current major version receives security fixes.

There aren't any published security advisories