The Agent Skills Directory

[COMMAND_EXECUTION]: The skill invokes a local script scripts/campaign-tracker.py using the host's Python environment to manage marketing campaign data. This execution happens within the Brand Context phase of the workflow.
[CREDENTIALS_UNSAFE]: The skill reads data from the user's home directory at ~/.claude-marketing/brands/, including brand profiles and guideline manifests. While these appear to be application-specific files, they involve accessing local data in the user's personal directory.
[INDIRECT_PROMPT_INJECTION]: The skill ingests and processes untrusted user-provided data such as brand identities, website URLs, and target queries during the discovery and optimization phases.
Ingestion points: External website content and brand identity details provided by the user (SKILL.md).
Boundary markers: None identified; instructions do not specify delimiters for external content.
Capability inventory: Local Python script execution and file read access in the home directory (SKILL.md).
Sanitization: No validation or sanitization of the input data is mentioned before it is processed by the agent.

aeo-geo