The Agent Skills Directory

Indirect Prompt Injection (LOW): The orchestration patterns ingest untrusted content from the local filesystem which could contain malicious instructions designed to influence the behavior of the orchestrated agents.\n
Ingestion points: SKILL.md identifies Read, Glob, and Grep as allowed tools for discovering and reading codebase content.\n
Boundary markers: Absent; the provided templates for sequential and parallel agent invocation do not include delimiters or instructions to ignore embedded commands in the data.\n
Capability inventory: The skill orchestrates multiple specialized agents (e.g., penetration-tester, security-auditor, backend-specialist) which work together to analyze and potentially modify codebase structures.\n
Sanitization: Absent; there is no mention of sanitizing, escaping, or validating file content before it is passed as context to subsequent agents in the chain.\n- No Executable Code (SAFE): The skill is composed entirely of markdown documentation and metadata. It does not include scripts, binaries, or configuration files that would be executed by the host environment.

parallel-agents