The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes shell commands using the GitHub CLI (gh) to perform tasks such as listing open issues, viewing issue content, and creating or editing issues and labels. These operations are central to the skill's functionality and are performed within a controlled orchestration pipeline.
[DATA_EXFILTRATION]: The skill reads local source code to determine impact areas and retrieves existing issue data from the GitHub API. This information is processed internally and then transmitted back to the GitHub platform to update or create issues. While these actions are intended, they represent the flow of local and external repository data across tool boundaries.
[PROMPT_INJECTION]: The skill processes external data from user reports and existing GitHub issue descriptions, which introduces a potential surface for indirect prompt injection. 1. Ingestion points: Untrusted data enters the system through user-provided bug/feature reports and existing issue bodies retrieved from the GitHub repository. 2. Boundary markers: The orchestration protocol separates the parent manager agent from specialized worker agents and uses local JSON files for intermediate data transfer, providing a degree of isolation. 3. Capability inventory: The skill has the capability to read local files, query the GitHub API, and perform authenticated writes (create/edit) to GitHub issues and labels. 4. Sanitization: Shell injection risks are mitigated by using the --body-file argument with the gh command, and all final issue drafts are presented to the user for explicit approval before submission.

issue-report-user