The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process data from external sources, which creates an attack surface where maliciously crafted instructions in those sources could influence the agent's behavior during the drafting or testing phases.
Ingestion points: Stage 1 of the workflow in SKILL.md explicitly prompts the agent to gather context by reading file uploads, links to shared documents (Google Drive/SharePoint), and team messaging channels (Slack/Teams).
Boundary markers: The instructions do not define delimiters or provide specific prompts to the agent to ignore instructions that might be embedded within the external context data.
Capability inventory: The skill uses file system manipulation tools (create_file and str_replace) to build the document and can invoke sub-agents to perform "Reader Testing."
Sanitization: There is no mention of sanitizing, escaping, or validating the external content before it is interpolated into the drafting process or passed to sub-agents for testing.

doc-coauthoring