The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes external, potentially untrusted data from GitHub and the web, which creates an attack surface for indirect prompt injection.
Ingestion points: The skill ingests data via gh api (issue templates), gh pr view (PR descriptions and diffs), WebFetch (external web pages), and MCP search tools (published documentation).
Boundary markers: The skill includes a critical safety constraint: "Never create, edit, or comment on an issue without explicit approval of the final draft," requiring a human-in-the-loop before execution.
Capability inventory: The skill possesses write capabilities through gh issue create, Write (to local scratch files), and Bash execution for GitHub CLI commands.
Sanitization: Instructions require the agent to restructure technical implementation details into user-facing behavior, which involves active processing and filtering of the input data rather than blind reflection.
[COMMAND_EXECUTION]: The skill relies on the GitHub CLI (gh) to interact with the environment and manage documentation workflows.
Evidence: The skill uses a variety of gh commands, including gh api for fetching repository content, gh issue list and gh search code for duplicate checks, and gh issue create for the final submission. These commands are scoped to the primary purpose of the skill but involve significant interaction with remote repository states.

docs-file-issue