The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill instructions include an installation command npx clawhub@latest install gog, which downloads and executes the 'clawhub' package from the npm registry.
[COMMAND_EXECUTION]: The documentation provides a shell command for skill installation using npx.
[DATA_EXFILTRATION]: The skill implements capabilities to read sensitive user information from Gmail and Drive and has the ability to send emails (send_gmail_message) or update documents. While this is the intended functionality, it creates a pathway for data to be moved out of the user's direct control if the agent is misled.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it ingests and processes untrusted data from external sources.
Ingestion points: Untrusted data enters the agent context via list_gmail_messages and search_gmail_messages in SKILL.md.
Boundary markers: The instructions lack explicit delimiters or safety warnings to ignore instructions embedded within the emails being processed.
Capability inventory: The skill possesses capabilities for send_gmail_message, create_google_drive_file, and update_google_docs_content as described in SKILL.md.
Sanitization: There is no evidence of sanitization or filtering of the content retrieved from external Google services before it is processed by the LLM.

gog