The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from multiple sources which could contain malicious instructions designed to influence the orchestrator agent's behavior.
Ingestion points: Reads implementation results from delegated workers located in the worker-results/ directory and monitors external feedback such as code review bot comments and human reviews on pull requests.
Boundary markers: Absent. The instructions do not specify the use of delimiters or 'ignore' instructions when reading worker outputs or PR comments.
Capability inventory: The skill has the authority to execute repository validation tools, manage git branches, and interact with pull request APIs.
Sanitization: Absent. There is no mention of filtering, escaping, or validating the content retrieved from external review monitors or worker results before processing.
[EXTERNAL_DOWNLOADS]: The skill explicitly suggests that the user install an external skill from a third-party GitHub user (DietrichGebert/ponytail) using the npx skills add command if the dependency is not met.
[COMMAND_EXECUTION]: The orchestrator is instructed to execute various shell-level operations, including repository-required formatting, linting, type checks, test suites, and build processes, as well as git commands for branch management and pull request creation.

ship-it