Security audit
Security checks for vulnerabilities and agentic risk
This skill is not malicious, but its selection rules push agents to invoke other skills even when the fit is unclear.
Review before installing in any environment with sensitive or side-effecting skills. The package has no code payload, but its rules can make an agent choose tools too eagerly; use it only if your orchestrator adds confirmation, risk checks, and a fail-closed path for ambiguous requests.
The rule 'if in doubt, use it' creates an unsafe default that favors invocation over restraint, which can cause unrelated or higher-risk skills to be executed on ambiguous matches. In a skill-selection system, this increases the chance of privilege overreach, unintended actions, and exploitation through prompt phrasing or keyword stuffing that nudges the agent into selecting a tool it should not use.
Referenced artifact was not completely inspected
3. **When unsure, consult SKILL.md**: if still unclear, read the candidate skill's `SKILL.md` to confirm fit and constraints.
Referenced artifact was not completely inspected
3. **When unsure, consult SKILL.md**: if still unclear, read the candidate skill's `SKILL.md` to confirm fit and constraints.
Referenced artifact was not completely inspected
3. **When unsure, consult SKILL.md**: if still unclear, read the candidate skill's `SKILL.md` to confirm fit and constraints.
Referenced artifact was not completely inspected
3. **When unsure, consult SKILL.md**: if still unclear, read the candidate skill's `SKILL.md` to confirm fit and constraints.
The rule to proceed with a skill "if in doubt, use it" creates an unsafe default that favors action over verification. In a skill-invocation framework, this can route ambiguous requests to the wrong capability, potentially triggering unintended data access, side effects, or execution paths without sufficient applicability checks.
The keyword matching design relies on simple trigger appearance and broad specificity heuristics, but defines no negative matches, domain boundaries, or safety constraints. This makes it easy for unrelated or adversarially phrased requests to match sensitive skills, especially in an orchestration layer where selecting the wrong skill can cause downstream harmful actions.
The judgment checklist ends with an ambiguous default that still instructs invocation when the system cannot determine fitness. In context, this skill governs selection for other skills, so an uncertainty-to-action bias amplifies risk across the whole environment by normalizing unsafe invocation under ambiguity.
No suspicious patterns detected.