The Agent Skills Directory

[PROMPT_INJECTION]: The skill presents an attack surface for indirect prompt injection by processing external, unverified data.
Ingestion points: The skill ingests untrusted data from the execution environment via the gitops_pod_log and gitops_app_event resource types in SKILL.md.
Boundary markers: There are no explicit delimiters or instructions provided to the agent to treat content from logs or events as data rather than instructions.
Capability inventory: The skill has access to the harness_execute tool, which allows it to perform actions like 'sync' on GitOps applications, creating a path for an attacker to influence infrastructure state.
Sanitization: The skill lacks sanitization or validation logic for the content retrieved from remote pod logs or events before it enters the agent's context.

gitops-status