The Agent Skills Directory

[SAFE]: The skill consists entirely of instructional markdown content designed to guide the agent's reasoning process. It does not contain any executable scripts, binary files, or hardcoded credentials.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to process untrusted external data, specifically source code, package configurations, and documentation from third-party repositories.
Ingestion points: Documented in SKILL.md as "Repo-level docs," "package/config files," "source code," and "official documentation."
Boundary markers: The skill instructs the agent to classify the question first and explicitly separate "Facts" from "Inferences" and "Unknowns," which provides a logical framework to mitigate the impact of malicious content embedded in data.
Capability inventory: The skill file itself contains no subprocess calls, network operations, or file-writing logic; it provides instruction only.
Sanitization: No explicit sanitization of input data is defined, but the requirement for cross-verification against primary sources serves as a behavioral safeguard.

study-repo