The Agent Skills Directory

[PROMPT_INJECTION]: The skill exhibits an attack surface for indirect prompt injection by processing untrusted data to perform operational actions on CI/CD pipelines.
Ingestion points: User-supplied parameters such as query in harness_search, search_term in harness_list, and resource_id in harness_execute (SKILL.md).
Boundary markers: No delimiters or "ignore embedded instructions" warnings are utilized when interpolating data into tool calls or reporting status back to the user.
Capability inventory: The skill can trigger, retry, interrupt, and approve or reject pipeline executions via the harness_execute MCP tool (SKILL.md).
Sanitization: The instructions do not define validation or escaping procedures for external content before it is processed by the agent or the underlying tools.

run-pipeline