The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes external data from the GET /contx-iq/v1/whoami endpoint, creating a potential surface for indirect prompt injection if the API response is manipulated. \n
Ingestion points: The agent ingests type and id strings from the IndyKite REST API response as described in SKILL.md and references/whoami-reference.md. \n
Boundary markers: SKILL.md and references/whoami-reference.md provide clear instructions to treat these fields as plain identifiers and never as instructions or commands to follow. \n
Capability inventory: The skill uses curl for network requests and instructs the agent to use the resulting data in subsequent AuthZEN or CIQ decision requests. \n
Sanitization: There is no programmatic sanitization of the API response in the scripts/whoami.sh script; protection relies on the agent following the documented safety constraints. \n- [COMMAND_EXECUTION]: The skill includes a bash script (scripts/whoami.sh) that executes curl commands to interact with IndyKite services. \n
The script includes a host-validation check that pins the destination to https://eu.api.indykite.com or https://us.api.indykite.com, ensuring credentials are only sent to authorized IndyKite domains. \n
Sensitive credentials and user tokens are handled via environment variables (API_KEY, BEARER_TOKEN) rather than being hardcoded in the skill code, following industry best practices for secret management.

indykite-ciq-whoami