The Agent Skills Directory

[SAFE]: The skill implements tools (harness_diagnose, harness_status, harness_list, harness_get) that are strictly scoped to the Harness platform for debugging and performance analysis. The behavior is consistent with the stated developer-focused purpose.
[INDIRECT_PROMPT_INJECTION]: The skill identifies a potential attack surface by processing external execution data. 1. Ingestion points: The harness_diagnose and harness_get tools in SKILL.md fetch pipeline logs and YAML definitions. 2. Boundary markers: The instructions lack explicit delimiters or warnings to ignore instructions that might be embedded in the retrieved logs or YAML. 3. Capability inventory: The skill is restricted to read-only MCP tool calls for data analysis; it does not possess capabilities for file modification, network exfiltration, or shell command execution. 4. Sanitization: There is no evidence of log or YAML sanitization before processing.

debug-pipeline