The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill fetches and processes content from GitHub issue bodies, which are external, user-controlled sources.
Ingestion points: SKILL.md uses gh issue view to retrieve issue data including the body.
Boundary markers: The skill instructions include a specific 'Security note' that explicitly directs the agent to treat the issue body as data only and to ignore any action directives or commands found within it.
Capability inventory: The skill utilizes git and gh CLI tools for branch management and issue creation.
Sanitization: The instructions mandate that the agent should skip and note anomalies to the user if the fetched content appears to be an instruction to the AI.
[COMMAND_EXECUTION]: The skill relies on shell commands for its core functionality, specifically git and gh CLI tools. It interpolates user-provided data such as issue titles and descriptions into command arguments for git checkout -b and gh issue create. While standard for such tools, this represents a potential surface for argument injection if the underlying LLM fails to properly quote the input.

start