The Agent Skills Directory

[COMMAND_EXECUTION]: The skill is designed to execute a local Python script named campaign-tracker.py to retrieve brand-specific campaign data. Running local scripts involves a level of risk regarding the integrity of the executable and its environment.
[PROMPT_INJECTION]: The skill processes untrusted input ('Original content') from external sources such as URLs or documents and uses this data to generate derivative content. This creates a vulnerability to indirect prompt injection. \n
Ingestion points: Source material provided via the 'Original content' field in SKILL.md. \n
Boundary markers: The skill lacks explicit delimiters or instructions to ignore potential injections within the source content. \n
Capability inventory: Includes local file reads and subprocess execution via Python. \n
Sanitization: No evidence of content validation or escaping is present in the instruction set.
[DATA_EXFILTRATION]: The skill accesses configuration and profile files located in the user's home directory (~/.claude-marketing/). While this appears to be the application's own data store, accessing sensitive files outside of a sandboxed project directory is a potential data exposure risk.

content-repurpose