Axiom is a public-beta workflow router. It adds instruction and hook surfaces to Codex and Claude Code, so a changed manifest, hook, Skill, or trust boundary can be security-relevant. A routing-quality bug is not automatically a security vulnerability.
Security reports are evaluated against the latest published release and the current main branch. Older releases are not actively maintained, and a reporter may be asked to reproduce an issue on the latest release. Fixes are not promised for every older version, and this project does not publish a response time or remediation SLA.
Do not publish exploitable details, secrets, proof-of-concept payloads, or sensitive repository data in an Issue, Discussion, pull request, or public comment.
GitHub private vulnerability reporting was verified as enabled on 2026-10-24. Use the repository's Report a vulnerability form to create a private security advisory. Do not open a public placeholder Issue for a vulnerability.
Include:
- the affected Axiom version, tag, or immutable commit;
- the affected host and version;
- the smallest safe reproduction;
- expected and observed trust-boundary behavior;
- impact and preconditions;
- whether a mutation, credential, external effect, or persistent change was involved; and
- a suggested correction, if known.
Use the public routing-case form for a false positive, false negative, ambiguity, or expected routing case that does not expose a vulnerability. Use the compatibility form for a fresh-session host observation. Sanitize every request and attachment.
If a routing case caused or could plausibly cause unauthorized credential use, external publication, destructive retention, unsafe Git publication, a persistent change outside the bound target, or execution of a changed hook, treat it as potentially security-sensitive and do not disclose the details publicly.
Read the Trust Model before assessing impact. In particular:
- selecting a route loads instructions; it never grants action authority;
- installed hooks must be compared with the exact checked-in definitions;
- host, operating-system, repository, credential, and connected-service permissions remain outside Axiom's control;
- request acceptance and command exit status are not proof of the intended outcome; and
- missing or unavailable evidence must not be reported as a pass.
Axiom is not a sandbox, policy-enforcement kernel, credential broker, malware detector, or guarantee that an agent, host, dependency, or external service is correct. It does not prevent every error or unsafe instruction. See the field-validation protocol for evidence labels and safe reproduction prompts.
Pull-request validation and release provenance are separate trust domains. Contributor checks validate a proposed merge tree with read-only repository permissions and without repository secrets. A passing check does not establish release provenance or authorize publication.
Release provenance depends on protected history, immutable release subjects, and independently verified publication evidence. Workflow detection cannot replace server-side prevention of unsigned protected updates or mutable release tags. A reusable commit-level check is evidence about that commit, not authorization for an exact tag-creation request.
Repository Governance owns the workflow trust boundaries, required checks, and dated ruleset snapshot. Its release-tag policy and manual re-verification procedure define what to inspect before relying on live remote state. See Release Documentation And Evidence for the distinction between candidate records and final publication evidence.