The Agent Skills Directory

[SAFE]: The skill provides a structured framework for complex implementation planning without introducing security vulnerabilities. It relies on standard internal tools for file exploration and sub-agent orchestration.
[PROMPT_INJECTION]: The skill contains an indirect prompt injection surface as it ingests codebase content to build implementation plans. This is a common characteristic of developer tools that process external files.
Ingestion points: Codebase files are read using Read, Grep, and Glob tools in Phase 0; user requests are captured and passed to sub-agents in Phase 1.
Boundary markers: The skill structures the sub-agent prompt using clear section headers such as 'fact summary block' and 'user's request', though it does not use explicit cryptographic delimiters or instructions to ignore embedded commands within that data.
Capability inventory: The analysis agents are provided with Read, Grep, and Glob tools for exploration and are explicitly instructed not to edit any files.
Sanitization: No explicit sanitization or escaping of ingested file content is performed prior to interpolation into the sub-agent prompt.

diverse-plan