The Agent Skills Directory

[COMMAND_EXECUTION]: The skill coordinates the execution of shell commands through a dedicated subagent to perform validation and testing.
Evidence: In SKILL.md, the consistency-qa agent is tasked with running npm run validate, npm test, and npm run evals.
[INDIRECT_PROMPT_INJECTION]: The skill processes repository content and user-provided instructions to drive a pipeline with significant system capabilities, presenting an indirect injection surface.
Ingestion points: Maintainer requests and repository files within the skills/ and evals/ directories.
Boundary markers: No specific delimiters or safety warnings for processing external content are defined in the instructions.
Capability inventory: The pipeline includes subagents capable of file modification, repository manifest updates, git operations (commit, tag, push), and shell command execution.
Sanitization: There is no evidence of content sanitization or validation before the data is passed to subagents or used to influence the pipeline flow.
[DYNAMIC_EXECUTION]: The orchestrator dynamically determines its execution path and spawns subagents based on the classification of incoming requests.
Evidence: Phase 1 of the pipeline describes classifying the maintainer's request to 'select stages' and subsequently 'spawn each selected stage as a subagent'.

brooks-harness