The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes shell commands using gh (GitHub CLI) and git to manage pull request state, metadata, and branches. It also executes a sequence of validation commands defined in the project's local .ai/agentic.config.json file to ensure code quality before completion.
[EXTERNAL_DOWNLOADS]: The skill invokes standard package managers such as npm, pnpm, bun, or cargo to install project dependencies within an isolated worktree. This behavior is triggered by the presence of standard lockfiles in the repository.
[PROMPT_INJECTION]: The skill processes untrusted input from Pull Request bodies and external tracking plans to determine the resume point of its workflow. This represents an Indirect Prompt Injection surface. However, the skill mitigates this by using structured parsing (looking for specific markdown headers and line prefixes) and establishing a rule that local project configuration and security checks take precedence over any instructions found in external references.

auto-continue-pr