The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to execute a local script using the command python campaign-tracker.py --brand {slug} --action list-campaigns. Since the script file is not provided in the skill payload, its functionality and security cannot be verified, posing a risk of arbitrary command execution if the script is malicious or improperly secured.
[DATA_EXFILTRATION]: The skill accesses several files within the ~/.claude-marketing/ directory, which is a hidden folder in the user's home directory. These files, such as profile.json and various brand guidelines, are considered sensitive as they may contain proprietary business data or user-specific configurations.
[PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection through the ingestion of external brand configuration data.
Ingestion points: Files located at ~/.claude-marketing/brands/{slug}/guidelines/ (e.g., restrictions.md, messaging.md).
Boundary markers: The skill lacks explicit delimiters or instructions to treat the content of these files as data rather than instructions.
Capability inventory: The skill has the capability to read local files and execute system commands via the Python interpreter.
Sanitization: There is no evidence of content sanitization or validation for the brand-specific markdown and JSON files before they are processed by the agent.

growth-engineering