context-gathering
Warn
Audited by Gen Agent Trust Hub on Sep 17, 2026
Risk Level: MEDIUMCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill explicitly directs the agent to use the '-k' (or --insecure) flag with curl when fetching Sitecore API JSON. This instruction deliberately bypasses SSL/TLS certificate validation, exposing the process to Man-in-the-Middle (MitM) attacks and allowing connections to potentially malicious endpoints.
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from JIRA User stories to extract endpoints and identifiers which are then used as inputs for fetching tools. It lacks necessary safeguards to prevent malicious instructions embedded in the data from being followed.
- Ingestion points: JIRA User story content (read once to extract Sitecore, BFF, and Figma identifiers).
- Boundary markers: Absent. There are no delimiters or instructions provided to the agent to treat the story content as data only.
- Capability inventory: File system write access, 'http' tool (curl execution), 'open-api-spec' MCP tool, and 'Figma' MCP server.
- Sanitization: Absent. The extracted URLs and operation IDs are used directly without validation or escaping.
Audit Metadata