indykite-capture-upsert-relationships

Pass

Audited by Gen Agent Trust Hub on Oct 13, 2026

Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill includes a bash script, scripts/capture.sh, which uses curl to transmit data to IndyKite's regional API endpoints.
  • Evidence: The script constructs and executes a curl command using provided environment variables (API_URL, API_KEY) and the generated JSON body.
  • Mitigations: The script implements a security check that restricts the API_URL to known IndyKite hosts (https://eu.api.indykite.com or https://us.api.indykite.com) before attaching the API_KEY header, protecting against credential leakage to unauthorized domains.
  • [INDIRECT_PROMPT_INJECTION]: The skill ingests user-supplied entity identifiers and relationship types to build a JSON payload, representing a potential surface for indirect prompt injection.
  • Ingestion points: User input defining source, target, and properties fields in SKILL.md.
  • Capability inventory: The agent has the ability to execute network requests via scripts/capture.sh.
  • Boundary markers: Instructions direct the agent to produce a structured JSON file.
  • Sanitization: The scripts/capture.sh script utilizes jq to validate that the input is a well-formed JSON object before transmission, reducing the risk of malformed or malicious payload execution.
Audit Metadata
Risk Level
SAFE
Analyzed
Oct 13, 2026, 03:29 PM