The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes external datasets including JSON snapshots and identity provider metadata. This creates an attack surface where data collected from third-party sources could contain instructions designed to influence the agent's behavior during triage.
Ingestion points: Workflow steps in SKILL.md involve analyzing JSON snapshots and GitHound samples located in the ../../references/ directory.
Boundary markers: There are no explicit markers or instructions provided to separate the external data from the skill's instructions or to treat the ingested data as untrusted.
Capability inventory: The skill body and referenced metadata do not define any automated scripts, network operations, or file system write capabilities.
Sanitization: The instructions recommend that the agent "adapt parameters safely" when constructing queries, but do not provide specific sanitization or validation logic for the external input.

openhound-github