The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses the GitHub CLI (gh) and git to interact with remote repositories and manage local code changes.
In scripts/fetch_issue.py, the skill executes the gh tool via subprocess.run to retrieve issue metadata, comments, and GraphQL timeline data. These calls use structured argument lists, which is a secure implementation.
The SKILL.md and reference files instruct the agent to use git commands for branching, committing, and bisection.
[INDIRECT_PROMPT_INJECTION]: The skill's primary function is to process and act upon user-contributed content from GitHub issues, which represents an indirect prompt injection surface.
Ingestion points: The scripts/fetch_issue.py script retrieves untrusted text from GitHub issue bodies and comments.
Boundary markers: The fetching script organizes the retrieved data into distinct sections (e.g., ── comments, ── reproduction leads). Both the SKILL.md instructions and the script's documentation provide explicit, high-priority warnings to treat this content as data rather than instructions to be followed.
Capability inventory: The skill has significant capabilities, including the ability to run code for reproduction, modify the repository using git, and create pull requests using gh.
Sanitization: The skill relies on the agent's adherence to the safety instructions rather than performing automated sanitization of the ingested issue text.

fix-issue