The Agent Skills Directory

[SAFE]: The skill is purely instructional and generative, focusing on the creation of Markdown documentation. It lacks tools for network communication (no curl, wget, or fetch) and shell execution, as explicitly noted in the instructions. Its filesystem operations are restricted to reading the codebase for context and writing to the docs/plans/ directory.- [INDIRECT_PROMPT_INJECTION]: The skill has a surface for indirect prompt injection by ingesting data from repository files and user input to generate plans.
Ingestion points: Files read via Read, Grep, and Glob tools, and user responses via AskUserQuestion.
Boundary markers: The skill does not define specific delimiters for untrusted content but produces structured Markdown output.
Capability inventory: The agent can Write and Edit local files.
Sanitization: No specific sanitization logic is provided.
Risk Assessment: The risk is low because the skill is restricted to 'planning only' and is explicitly forbidden from writing executable application code or creating external issues.

plankit