The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses curl and jq within the scripts/read-schema.sh script to interact with the IndyKite Data Schema API and process JSON responses. These are standard system utilities used for the skill's primary functionality.
[DATA_EXFILTRATION]: The skill transmits a user-provided API_KEY (via environment variable) to IndyKite's regional API endpoints (eu.api.indykite.com or us.api.indykite.com). The scripts/read-schema.sh script implements host pinning, which validates the API_URL against a whitelist of known IndyKite hosts before sending the credential, preventing accidental or malicious exfiltration to unauthorized domains.
[INDIRECT_PROMPT_INJECTION]: The skill represents a surface for indirect prompt injection as it ingests data from an external API (GET /data-schema/v1/) that is then processed by the agent using jq recipes.
Ingestion points: JSON response from the IndyKite Data Schema API.
Capability inventory: Network requests via curl and data processing via jq in scripts/read-schema.sh and references/data-schema-reference.md.
Boundary markers: None identified in the prompt templates.
Sanitization: The skill uses strict host pinning in the shell script to ensure data is only pulled from trusted vendor infrastructure. The ingested data is limited to schema-level metadata (node/relationship types and counts), which presents a minimal risk to agent logic compared to raw user content.

indykite-data-schema