The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes shell commands using gh (GitHub CLI) and git to manage repository data and issues. These commands are legitimate and limited to the skill's stated purpose of filing bug reports.
[EXTERNAL_DOWNLOADS]: The skill interacts with the official GitHub API and repository endpoints to retrieve metadata and post issues. These operations target well-known and trusted services.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection due to the processing of untrusted user input to generate issue drafts. 1. Ingestion points: User-provided bug descriptions and stack traces in Phase 2. 2. Boundary markers: Absent (user content is interpolated directly into a markdown template). 3. Capability inventory: Ability to create issues and labels via the Bash tool. 4. Sanitization: None provided in the instructions. Note: The risk of malicious instruction execution is significantly reduced by the mandatory human-in-the-loop confirmation required in Phase 3 before any state-changing command is executed.

bug