Security audit
Security checks for vulnerabilities and agentic risk
This is mostly a cloud-architecture advice skill, but it asks for shell execution access and has mismatched trigger instructions that could activate it outside its stated purpose.
Install only if you are comfortable with a cloud-advice skill receiving shell execution permission. Prefer a version that removes exec access and corrects the trigger section to cloud architecture tasks only.
The trigger section defines broad, unrelated creative-design scenarios for invoking a cloud architecture skill, making routing boundaries unclear. Poorly scoped triggers are dangerous because they can cause the agent to load a skill with unnecessary permissions or irrelevant behavior in contexts where it should never run.
Granting exec capability to a skill whose stated purpose is generating cloud architecture recommendations violates least privilege. Even without explicit malicious commands, the extra capability expands the attack surface if the skill is invoked under adversarial prompting or combined with conflicting instructions elsewhere in the file.
The capability coverage keywords are expansive, fragmented, and not bounded by precise trigger rules or exclusions. In context, this makes the skill easier to activate on loosely related prompts and increases the chance of overreach, especially since the skill also advertises unnecessary exec access.
The trigger condition says the skill should be used for creative design, UI, posters, and branding, which directly contradicts the rest of the document describing a cloud architecture advisory skill. Ambiguous or incorrect invocation guidance can cause the agent to activate this skill in unrelated contexts, increasing the chance of inappropriate tool access and misleading outputs.
The skill claims it is pure Markdown, does not require exec, and does not connect to cloud APIs, yet the metadata allows exec and the error handling instructs the agent to run ping. This inconsistency weakens operator expectations and may normalize unnecessary command execution in a skill that should be advisory-only.
No suspicious patterns detected.
