The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill manages asset permissions based on data retrieved from external tools, creating a surface where malicious metadata or asset names could influence agent behavior.
Ingestion points: The skill ingests untrusted data via search_media_library (user-provided names) and tool outputs from build_access_list as described in SKILL.md.
Boundary markers: No specific delimiters or instructions to ignore embedded content are provided for the data retrieved from search or ACL tools.
Capability inventory: The skill possesses sensitive permission-modifying capabilities through the update_asset_acl and update_media_collection tools.
Sanitization: The instructions include a mitigation strategy by requiring the agent to use system-validated userId and groupId values from official tool responses rather than user-supplied strings.

asset-access-control