The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes user-provided data such as product requirements and market research notes to populate templates. While this constitutes an attack surface, the skill serves as a static reference for the agent to follow, and the templates include structured fields that help organize data.
Ingestion points: User-provided text in queries and referenced product documentation files.
Boundary markers: None specified in the templates, though the use of structured JSON templates in references/output-templates.md provides some implicit structure.
Capability inventory: The skill uses Read, Glob, Grep, WebFetch, and WebSearch tools. No high-risk tools like code execution or shell access are requested in the allowed-tools list.
Sanitization: No explicit sanitization of user input is described.
[COMMAND_EXECUTION]: Several reference files (market-competitive.md, competitive-analysis-guide.md) contain examples of using the gh (GitHub CLI) to gather market intelligence. These are provided as instructional patterns for the agent to follow during research tasks and are not obfuscated or autonomous injection attempts.
[DYNAMIC_EXECUTION]: The rules/business-roi.md file includes Python code snippets for calculating financial metrics such as NPV and IRR using the scipy library. These are benign, domain-relevant code examples intended to assist the agent in performing mathematical calculations requested by the user.

product-frameworks