The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to retrieve and process large amounts of data from external web sources through search and fetch tools. This content is stored in local workspace files (materials.md) and subsequently used to synthesize a final report. This creates a surface for indirect prompt injection where malicious instructions embedded in web pages could influence the agent's behavior.
Ingestion points: External data is ingested via search results and page fetching during Stage 5 (Evidence Collection), as defined in SKILL.md and reference/methodology.md.
Boundary markers: The instructions do not specify the use of clear delimiters (like XML tags or triple quotes) or 'ignore embedded instructions' warnings when writing fetched content into the materials.md file.
Capability inventory: The skill has the capability to write and update files in the local workspace (research/ directory) and utilize search/fetch tools.
Sanitization: There is no requirement for the agent to sanitize, filter, or escape the content retrieved from external URLs before it is stored or synthesized.

deep-research