The Agent Skills Directory

[COMMAND_EXECUTION]: The skill makes extensive use of the GitHub CLI (gh) to perform repository management tasks, including listing issues, creating labels, creating new issues, and closing existing ones. This behavior is directly aligned with the skill's primary purpose.
[INDIRECT_PROMPT_INJECTION]: The skill ingests potentially untrusted data from the titles and bodies of GitHub issues during the 'Gather' and 'Explore' phases. This data could contain instructions designed to manipulate the agent's logic.
Ingestion points: Step 1 in SKILL.md uses gh issue list to fetch issue numbers, titles, and bodies into the agent's context.
Boundary markers: The instructions do not specify any delimiters or safety prompts to isolate issue content from the agent's instructions.
Capability inventory: The skill has the capability to write to the repository via gh issue create, gh issue edit, and gh label create (Step 6).
Sanitization: No specific sanitization or escaping of the ingested issue data is mentioned.
Mitigation: The skill includes a 'Confirm the batch' step (Step 5) which requires human review of the generated tickets and labels before the final 'File' step is executed, significantly reducing the risk of unauthorized actions.

sweep