The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes several standard command-line tools, including git and gh (GitHub CLI), to detect repository information, generate diffs, manage branches, and create issues. These operations are aligned with the skill's stated purpose of managing GitHub issues.
[DATA_EXFILTRATION]: The skill implements a feature to collect local workspace changes (staged, unstaged, and untracked files) and upload them to GitHub by either embedding them in the issue body or pushing them to a remote branch. While this involves sending local code to an external service, it is the intended functionality designed to provide context for autonomous development agents. The risk of accidental data exposure is mitigated by mandatory user confirmation steps before any data is transmitted.
[INDIRECT_PROMPT_INJECTION]: The skill processes content from the local workspace (file contents and diffs) which could theoretically contain malicious instructions designed to influence the LLM during the issue drafting phase. However, the final draft is presented to the user for explicit review and confirmation before the issue is created, providing a human-in-the-loop defense against such attacks.
Ingestion points: Workspace changes are read using git diff and cat (via references/workspace-changes.md).
Boundary markers: The instructions encourage writing specific templates, but do not explicitly use guardrail markers for the ingested diff data.
Capability inventory: Uses gh issue create and git push to send data to GitHub.
Sanitization: No specific sanitization of file contents is performed prior to inclusion in the issue body.

create-issue