The Agent Skills Directory

Command Execution: The skill uses the GitHub CLI (gh) to retrieve issue data, edit issue bodies, and minimize comments via the GraphQL API. While these are standard operations for a repository maintenance tool, they involve executing shell commands with interpolated variables.
Indirect Prompt Injection Surface: The skill ingests untrusted data from GitHub issue bodies and comments (dotnet/machinelearning issue labeled repo-health). This data is processed to generate updates for the dashboard, creating a potential vector where malicious instructions embedded in comments could influence the agent's behavior.
Ingestion points: GitHub issue body and comments are retrieved using gh issue view in references/playbook.md.
Boundary markers: None explicitly defined in the provided methodology to delimit untrusted data from instructions.
Capability inventory: The skill has the ability to modify issue content (gh issue edit) and hide comments (gh api graphql mutation) as documented in references/playbook.md.
Sanitization: The skill implements a length floor (ratio must be at least 0.80) and a structure check (verifying required headers like # 🏥 Repo Health Dashboard) to prevent significant data loss or structural corruption, though it does not filter specific content types.

repo-health-groom