The Agent Skills Directory

Local Command Execution: The skill references validation commands using project-specific tools such as dotnet, python, and shell scripts located in the eng/ directory. These commands are intended for local development workflows and utilize vendor-owned resources, which is a standard practice for maintaining skill quality.
Indirect Prompt Injection Surface: The diagnostic process requires the agent to analyze external data sources, including evaluation PR comments, transcripts, and artifact files. While this is necessary for the skill's function, it introduces a potential surface where adversarial content in the data could influence the agent's behavior.
Ingestion points: The skill ingests data from evaluation results, PR comments, and trial transcripts (SKILL.md).
Boundary markers: No specific delimiters or 'ignore' instructions are provided to the agent for handling potentially untrusted segments within these inputs.
Capability inventory: The skill provides instructions for executing shell commands and local project scripts (SKILL.md).
Sanitization: The skill does not include instructions for filtering or sanitizing the data extracted from external artifacts before processing.

improve-skill-quality