The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes the GitHub CLI (gh) to perform repository management tasks, including issue creation, project board updates, and label manipulation. This is the intended and documented behavior of the skill.
[SAFE]: The skill implements a 'HARD-GATE' policy, explicitly instructing the agent not to create issues or assume credentials without user confirmation. This mitigation reduces the risk of automated or accidental data modification.
[SAFE]: All external tools (gh CLI) and API endpoints (GitHub GraphQL and REST APIs) belong to a well-known and trusted service provider (GitHub/Microsoft).
[INDIRECT_PROMPT_INJECTION]: The skill processes data from repository metadata, issue templates, and user inputs to generate GitHub content. This presents a potential surface for indirect injection (e.g., if a repository template contained malicious instructions). However, the risk is assessed as safe due to the mandatory human-in-the-loop confirmation step.
Ingestion points: Issue templates, project field options, and user-supplied issue content in SKILL.md and references/gh-commands.md.
Boundary markers: Not explicitly defined in the prompt instructions.
Capability inventory: Write access to GitHub issues, comments, labels, and project items via the gh CLI.
Sanitization: Relies on the standard parameter handling of the GitHub CLI.

github-issue