The Agent Skills Directory

[COMMAND_EXECUTION]: The skill provides numerous examples of shell commands involving the GitHub CLI (gh), along with utility tools like jq, base64, diff, and grep. These are used for remote repository analysis and local processing of fetched data.
[EXTERNAL_DOWNLOADS]: Several commands facilitate the retrieval of data from GitHub's servers, including repository contents, metadata, and search results. Documentation for gh extension install and gh repo clone also describes the process of downloading external code and extensions.
[CREDENTIALS_UNSAFE]: The skill includes documentation for gh auth subcommands, such as gh auth token and gh auth status --show-token, which can output authentication tokens to the terminal. These are standard administrative functions of the CLI tool.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to fetch and analyze content from remote repositories (e.g., package.json, source code, READMEs). This establishes an attack surface where an agent could ingest untrusted data containing malicious instructions. The skill provides the mechanics for fetching this data but does not provide specific boundary markers or sanitization logic.
Ingestion points: Remote file contents and metadata fetched via gh api and gh search (referenced in SKILL.md and remote-analysis.md).
Boundary markers: Absent in the provided command templates.
Capability inventory: Shell command execution via gh, jq, and others across multiple reference files.
Sanitization: Not explicitly addressed in the instructions.

gh-cli