convex
Fail
Audited by Snyk on Nov 20, 2026
Risk Level: CRITICAL
Full Analysis
CRITICAL E004: Prompt injection detected in skill instructions.
- Potential prompt injection detected (medium risk: 0.30). The skill is a Convex routing guide, but it has medium-risk behavior: it pulls remotely served procedure text that can change without a skill update, and one bundled skill sends the full coding-session transcript to the vendor, which may expose secrets the user didn't expect to leave their machine.
MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).
- Third-party content exposure detected (low risk: 0.10). The skill's required routing sends production errors and request logs to convex-monitor, convex-self-heal, and convex-advisor, which can contain attacker-influenced strings from public endpoints, but the agent must actively pull them and the served catalog itself is first-party.
MEDIUM W012: Unverifiable external dependency detected (runtime URL that controls agent).
- Potentially malicious external URL detected (high risk: 0.70). The skill directs the agent to fetch a capability catalog from a non-official Convex-looking host (basic-anteater-667.convex.site, a user-style deployment subdomain) and follow the served /capability/.md docs as procedure instructions, preferring them over bundled ones. This is an inbound instruction dependency from an unverified, arbitrary-deployment source, so it is weak provenance and high tier. The docs.convex.dev reference is only a documentation citation.
MEDIUM W013: Attempt to modify system services in skill instructions.
- Attempt to modify system services in skill instructions detected (low risk: 0.10). The skill routes to bundled and served procedures that change live Convex deployments, such as setting env vars, attaching custom domains and promoting schema migrations, but no supplied text establishes explicit irreversible operational loss, so gate 2 fails.
Issues (4)
E004
CRITICALPrompt injection detected in skill instructions.
W011
MEDIUMThird-party content exposure detected (indirect prompt injection risk).
W012
MEDIUMUnverifiable external dependency detected (runtime URL that controls agent).
W013
MEDIUMAttempt to modify system services in skill instructions.
Audit Metadata