The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill instructs the agent to process data from external API responses during live smoke checks, creating an attack surface where untrusted data could influence agent behavior.
Ingestion points: The shell snippet in SKILL.md uses curl to fetch data from $HEALTH_URL and pipes it to jq.
Boundary markers: There are no explicit instructions or delimiters provided to the agent to treat the API response as untrusted data.
Capability inventory: The skill has access to shell commands (curl, jq) and the Go toolchain for running tests.
Sanitization: The shell script validates specific JSON fields using jq, but no sanitization or validation of the fetched content is specified before it enters the agent's broader context.
[COMMAND_EXECUTION]: The skill contains a shell script example designed to be executed by the agent to verify service health.
Evidence: A bash script in SKILL.md uses curl and jq to validate the status and body of a response from a URL defined in the HEALTH_URL environment variable.

z-qa-api