The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructions direct the agent to execute a local Python script named campaign-tracker.py with arguments dynamically generated from the session context.\n
Evidence: Found in SKILL.md: python campaign-tracker.py --brand {slug} --action list-campaigns.\n
Risk: The {slug} parameter is interpolated directly from the brand context. If not properly sanitized, this could lead to command injection if a brand slug contains malicious shell characters.\n- [DATA_EXFILTRATION]: The skill is configured to read potentially sensitive marketing data and compliance rules from a hidden directory in the user's home folder.\n
Evidence: SKILL.md contains multiple references to paths such as ~/.claude-marketing/brands/{slug}/profile.json and ~/.claude-marketing/brands/{slug}/guidelines/_manifest.json.\n
Risk: This allows the agent to access user-specific configuration and marketing strategy files, which could be exposed if the agent is compromised.\n- [PROMPT_INJECTION]: The skill audit process involves ingesting data from external websites and server logs, which presents an attack surface for indirect prompt injection.\n
Ingestion points: Data enters the context from "Website URL", search console reports, and "Log File Analysis".\n
Boundary markers: Absent. The skill does not define clear delimiters or instructions to ignore embedded directives in the audited content.\n
Capability inventory: The skill has file system access to the home directory and the ability to execute local Python scripts via subprocesses.\n
Sanitization: Absent. There are no instructions for validating or sanitizing the content retrieved from external URLs or log files.

technical-seo