The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill provides instructions for downloading and parsing raw telemetry data from external URLs via the query_result_json metadata field. This represents a vulnerability surface where malicious content embedded in logs or traces could be processed by the agent.
Ingestion points: The query_result_json URL is explicitly recommended as a source for precise analysis in SKILL.md and result-interpretation.md.
Boundary markers: The guidance lacks instructions for implementing boundary markers or explicit warnings to the agent to ignore instructions embedded within the telemetry data.
Capability inventory: The skill assumes the agent can perform network requests (curl, urllib.request) and execute parsing tools (jq, Python scripts) on the retrieved data.
Sanitization: No sanitization or validation of the external JSON content is recommended before the agent parses and acts upon the values.
[SAFE]: The skill consists entirely of markdown documentation and JSON query examples. It does not contain any executable scripts, obfuscated payloads, or persistence mechanisms. All external references are directed toward the vendor's own official infrastructure.

query-patterns