The Agent Skills Directory

[COMMAND_EXECUTION]: The skill provides a set of templates for using the gh command-line tool. These commands are optimized for agentic consumption by using --json flags and --jq filtering to ensure structured and predictable data processing.
[PROMPT_INJECTION]: The skill identifies an attack surface for indirect prompt injection by facilitating the ingestion of untrusted data from external sources (GitHub PRs and Issues).
Ingestion points: The gh pr view and gh issue view commands are configured to retrieve the body field from pull requests and issues (SKILL.md, lines 40, 107).
Boundary markers: The templates do not implement delimiters or instructions to treat external data as potentially untrusted.
Capability inventory: The skill provides the agent with write capabilities such as gh workflow run, gh issue transfer, and gh issue develop (SKILL.md, lines 89, 152, 161).
Sanitization: No explicit sanitization or validation of the retrieved content is described in the workflow patterns.
[COMMAND_EXECUTION]: The skill demonstrates the use of dynamic context injection syntax (!command). The documented examples are used for read-only status reporting (gh pr checks and gh run view) and adhere to platform-specific error handling best practices.

gh-cli-agentic