The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it ingests data from external spreadsheets that may contain adversarial instructions.
Ingestion points: The actions read, list_notes, and get_headers (defined in SKILL.md and schema.md) allow the agent to fetch arbitrary content from spreadsheet cells and notes into its context.
Boundary markers: The skill instructions do not specify or require the use of delimiters (like XML tags or triple quotes) or "ignore embedded instructions" warnings when processing spreadsheet data.
Capability inventory: The agent has high-impact capabilities including update_row, delete_rows, share, and export_sheet across all scripts.
Sanitization: There is no evidence of content sanitization, escaping, or validation performed on the spreadsheet data before it is presented to the agent.
[DATA_EXFILTRATION]: The skill provides built-in mechanisms to move data from the Google Sheets environment to external locations.
The export_sheet action (SKILL.md) allows the agent to send spreadsheet data to the "AgentPMT File Manager" hosted on the vendor's infrastructure (agentpmt.com).
The share action (SKILL.md) allows granting spreadsheet access to arbitrary email addresses, which could be used to exfiltrate data to unauthorized third-party accounts.

google-sheets