The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill instructs the agent to ingest and analyze content from arbitrary external websites and local files, creating a surface for indirect prompt injection attacks.
Ingestion points: The Competitor Article Fetch Workflow section in SKILL.md directs the agent to fetch content from competitor URLs using mcp_web_fetch or WebSearch. The Report Workflow also specifies reading Excel/CSV files and the project-context.md file.
Boundary markers: The instructions do not define boundary markers (e.g., XML tags or delimiters) or explicit 'ignore embedded instructions' warnings for the data being fetched.
Capability inventory: The skill utilizes web fetching, web searching, and file system read capabilities to process external and local data.
Sanitization: There are no instructions for sanitizing or filtering the content retrieved from external sources before it is processed or included in the agent's output.

competitor-research