The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses the GitHub CLI (gh issue create) and GraphQL mutations to manage issues. This behavior is the primary purpose of the skill and is strictly gated by human-in-the-loop authorization.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted user input in the form of feature ideas and notes. Every skill that translates user requests into tool actions has this attack surface, but it is well-mitigated here.
Ingestion points: User-provided ideas or notes described in the skill's 'Propose Feature' goal (SKILL.md).
Boundary markers: The 'Review and file' section enforces a manual review step and requires the agent to 'Show the full draft' before filing.
Capability inventory: The skill has the capability to write to GitHub via gh issue create and GraphQL (SKILL.md).
Sanitization: The skill mandates following 'anonymization rules' and checking proposed commands against existing code or help files.
[SAFE]: The skill references project files and a vendor-owned GitHub repository (github.com/grafana/gcx) for context and examples, which is consistent with the author's identity and stated purpose.

propose-feature