The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes data from external sources (Harness pipeline configurations and execution logs) and possesses the capability to execute actions like triggering or aborting pipelines.
Ingestion points: Data enters the context via harness_get (pipeline details, execution status) and harness_list (input sets, approval instances).
Boundary markers: The instructions do not explicitly define delimiters for tool outputs, but they do include a "Safety" section requiring explicit user confirmation before high-stakes actions like production deployments.
Capability inventory: The skill can search, list, get details, execute, monitor, approve, retry, and interrupt Harness pipelines and executions across all scripts (SKILL.md).
Sanitization: No specific sanitization or filtering of external data is described within the skill instructions.

run-pipeline