The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructions in SKILL.md require the execution of a local Python script campaign-tracker.py with the parameter --brand {slug}. This dynamic parameter insertion poses a risk of command injection if the {slug} variable is not strictly validated.
[DATA_EXFILTRATION]: The agent is directed to read brand profiles and guidelines from the user's home directory (e.g., ~/.claude-marketing/brands/{slug}/profile.json). Accessing the ~/ path is a data exposure risk, though it appears necessary for the skill's stated marketing automation purpose.
[PROMPT_INJECTION]: The skill processes untrusted data from external sources (customer reviews), making it susceptible to indirect prompt injection. 1. Ingestion points: Customer feedback and sentiment data from external review platforms like Google, Yelp, and G2. 2. Boundary markers: The skill does not implement delimiters or ignore-instructions logic for these inputs. 3. Capability inventory: The skill can read local files and execute Python scripts. 4. Sanitization: No sanitization or validation of the external review content is mentioned in the skill instructions.

reputation-management