Security audit
Security checks for vulnerabilities and agentic risk
This skill is a disclosed personal-brand content workflow that stores local draft and performance files, with no evidence of hidden network access, credential handling, or destructive behavior.
Install this only in a workspace where you are comfortable with the agent reading and writing brand, audience, draft post, and performance-tracking files. Use explicit requests when invoking it, review generated drafts and memory updates before relying on them, and run the optional scripts from the intended project directory.
The skill advertises activation through very broad, natural-language requests like asking for help posting content, which can easily overlap with ordinary user conversation. In a file-system-aware agent environment, this increases the chance the skill will auto-engage unexpectedly and start steering workflow or causing reads/writes without sufficiently explicit user intent.
The trigger table uses vague conversational phrases such as 'Help me post 5 things today' and 'Analyze my recent performance' without strict activation syntax or scope constraints. Because these are common requests, the agent may invoke privileged skill behavior during unrelated sessions, potentially reading persistent files, modifying content logs, or applying behavior the user did not explicitly request.
The trigger phrases are broad enough that ordinary user requests like 'help me grow' or loosely related posting requests can activate the content-generation workflow unintentionally. Because activation causes mandatory file reads/writes and state transitions, a benign conversational prompt could trigger persistent memory updates or content logging the user did not explicitly intend.
The inline user-facing text states a concrete behavior: "The AI will pause and run the Initialization Checklist." However, the code in this branch only emits warning messages when required files are missing and then exits normally. This is an active contradiction between the script's documented/runtime messaging and its actual behavior.
No suspicious patterns detected.
