Security audit
Security checks for vulnerabilities and agentic risk
This skill is a local competitor-research assistant that saves research notes and preferences, with no executable code or evidence of credential use, exfiltration, or hidden behavior.
Install this only if you are comfortable with competitor research notes, preferences, and business context being saved locally under ~/competitor-research/. Avoid putting confidential credentials or private third-party system access into the workflow; the skill is intended for public competitive research and local note keeping.
The setup explicitly encourages the agent to proactively activate on broad triggers like 'competitors' or 'market research' within the first few exchanges. That creates an ambiguous invocation boundary where the skill may engage or steer conversations without a clear, current user request, increasing the risk of overreach and unintended collection or use of user context.
The skill instructs the agent to write the user's activation preference into persistent memory without any notice, consent flow, or retention limits. This is dangerous because it silently stores behavioral preferences across sessions, which can surprise users, expand profiling, and cause future actions based on data they did not realize would persist.
L141 says 'No config keys visible — use natural language, not "depth: quick"', but the file explicitly prescribes visible key-style fields such as 'status:', 'version:', 'last:', 'integration:', 'depth:', 'focus:', and 'style:' in the memory structure at L8-L23. This is an active contradiction between documented intent and the artifact the skill tells the developer to create.
No suspicious patterns detected.
