The Agent Skills Directory

[DATA_EXFILTRATION]: The skill requires access to sensitive local configuration files containing API keys and paths to service account private keys to function.
Evidence: The SKILL.md and references/auth-setup.md files document the use of ~/.config/claude-seo/google-api.json and service account JSON files. The underlying Python scripts (e.g., scripts/google_auth.py) are designed to read these credentials for authentication with Google APIs. While this is the intended primary purpose, it involves handling sensitive credential paths.
[INDIRECT_PROMPT_INJECTION]: The skill features commands that ingest and process external data from local files, which could contain malicious instructions designed to influence the agent's logic.
Ingestion points: SKILL.md defines commands like /seo google inspect-batch <file>, /seo google index-batch <file>, and /seo google report --data <json> which read from user-provided file paths.
Boundary markers: No explicit boundary markers or 'ignore' instructions are provided to the agent to distinguish between data and potential embedded instructions within these files.
Capability inventory: The skill possesses the capability to execute shell commands (via Python subprocess calls to utility scripts) and perform network requests to various Google API endpoints.
Sanitization: There is no documentation or evidence of input validation or sanitization for the content of the files processed by the batch or reporting commands.

seo-google