The Agent Skills Directory

[SAFE]: The skill is entirely instructional and serves as a reference for official GitHub CLI commands and standard GitHub Action workflows. It does not contain executable code beyond shell examples and YAML templates intended for user implementation. It focuses on documented tools and follows security best practices, such as using least-privilege permissions in GitHub Actions and advising against public disclosure of security vulnerabilities.
[PROMPT_INJECTION]: The skill identifies a potential surface for indirect prompt injection as the automation patterns it describes in references/automation-actions.md process user-generated issue content. It mitigates this risk by demonstrating validation techniques and structured triage workflows.
Ingestion points: Automation workflows like validate-template.yml and auto-assign-label.yml ingest data from context.payload.issue.body or issue.title into scripts.
Boundary markers: The validate-template.yml example shows how to verify the presence of specific Markdown headers (REQUIRED_SECTIONS) to ensure structured input.
Capability inventory: The skill uses subprocess for gh CLI commands and describes mutations via gh api and actions/github-script.
Sanitization: Examples include string-presence checks for template compliance and the use of triage labels to gate issue progression.

@tank/github-issues