The Agent Skills Directory

[SAFE]: The skill is designed for context engineering and requirement clarification. It contains explicit constraints that prevent the agent from writing production code, creating branches, committing, or pushing changes. These guardrails significantly limit the potential for malicious activity or accidental system modifications.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to process untrusted data from user requests and repository evidence to generate documentation. While it acts on external data, the risk is minimal because its capabilities are strictly restricted to writing markdown files in the docs/context-engineering/ directory, and it lacks tools for code execution or network exfiltration. The instructions include boundary-like rules, such as separating user statements from repository evidence and labeling inferences as 'proposed', which helps maintain data integrity.

intent