agent-rules-kit is a local diagnostic CLI for AI agent instruction files.

It is not a security scanner, provides no security guarantees, and must not be described as making a repository secure.

v0.4.0 is the current published GitHub Release and PyPI package.

v0.3.0 remains the previous published GitHub Release and PyPI package baseline.

The project is still maintained on a best-effort basis. There is no commercial SLA, no guaranteed response time, and no guarantee that every security-relevant issue will be found or fixed.

agent-rules-kit==0.4.0 is published on PyPI. Future PyPI availability claims must be verified per release before updating this policy.

The project must preserve these boundaries:

• read-only by default;
• no network access in runtime behavior;
• no LLM dependency in runtime behavior;
• no execution of commands from analyzed repositories;
• no unsupported security claims;
• no printing of raw secrets;
• no file modification during check;
• no file modification during init --dry-run;
• file modification only through explicit init --write user intent;
• existing root AGENTS.md must be backed up before replacement.

Potential secrets must be redacted before being shown in console, JSON, Markdown, logs, or tests.

Do not commit real secrets, tokens, credentials, cookies, private keys, private URLs, or customer data.

Use fake examples only.

Private vulnerability reporting is enabled for this repository.

Use GitHub private vulnerability reporting for sensitive vulnerability reports when available.

See docs/PRIVATE-VULNERABILITY-REPORTING.md for the verification record and limits of this disclosure channel.

If a sensitive issue cannot be reported privately through GitHub, do not publish secrets, exploit details, private URLs, customer data, or sensitive repository contents. Open only a minimal public issue requesting a private contact path.

For non-sensitive security boundary issues, open a GitHub issue with a minimal reproduction.

agent-rules-kit does not aim to:

• prove that a repository is secure;
• replace human security review;
• scan dependencies for vulnerabilities;
• validate CI/CD supply chain security;
• execute repository commands to confirm behavior;
• inspect private services, credentials, or infrastructure;
• provide complete secret scanning.

Security response is best-effort for the current published 0.4.x GitHub Release line.

There is no commercial SLA or guaranteed response time.

For future releases or broader public distribution, the maintainer should re-check and document:

• supported versions;
• expected response time;
• disclosure handling;
• whether GitHub Security Advisories or private vulnerability reporting are enabled;
• whether the published package channel changes the support policy.

Contributors and AI assistants must follow AGENTS.md.

Any change that touches secret detection, redaction, file traversal, write behavior, symlink handling, or command execution boundaries must be treated as security-sensitive and reviewed carefully.

If a change could expose secrets, execute untrusted code, write outside the intended project root, or make exaggerated security claims, stop and escalate before implementation.