The Agent Skills Directory

[PROMPT_INJECTION]: The skill retrieves untrusted data from GitHub issues using the 'gh issue view' command and embeds this content into prompts for sub-agents without sanitization or strict boundary markers. Ingestion points: Phases 3 and 4.5 in 'SKILL.md' use 'gh' to fetch issue body and requirements. Boundary markers: The agent prompt template in Phase 3 does not include delimiters or instructions to ignore embedded commands. Capability inventory: Spawned worker agents have access to the shell, git, and the GitHub CLI, allowing for code modification and repository interaction. Sanitization: No escaping or validation is performed on the issue content before interpolation.
[COMMAND_EXECUTION]: The skill documentation and reference files provide instructions for executing system-level commands and utilizing administrative CLI tools. Evidence: 'references/commands.md' and 'references/pitfalls.md' include commands for process inspection and termination such as 'ps aux' and 'kill '. Evidence: The skill makes extensive use of the 'gh' (GitHub) CLI, including 'gh api' calls to interact with the GitHub REST API and 'gh pr merge' for repository state changes.

blitz