The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to perform local filesystem operations using system shell commands.
Evidence: In SKILL.md, it states "必须用 bash 抽读相关作品原文做参照" (must use bash to read original works).
Technical details: rules/corpus-guide.md provides specific command patterns for the agent to execute, including ls, head -150, sed -n '1200,1350p', and grep -A 60.
Target paths: Commands target absolute local paths such as /Users/fred/Documents/重点文献/公有领域世界文学名著集/.
[PROMPT_INJECTION]: The skill demonstrates an indirect prompt injection surface due to the ingestion of external data without proper sanitization.
Ingestion points: Local text and markdown files are read directly into the agent's context using shell tools (defined in rules/corpus-guide.md).
Boundary markers: There are no instructions or delimiters provided to prevent the agent from obeying instructions that might be embedded within the processed literary works.
Capability inventory: The agent has access to shell execution (bash) and filesystem exploration tools.
Sanitization: No validation or filtering of the file content is performed before it is processed by the agent.
[COMMAND_EXECUTION]: The use of absolute hardcoded paths (/Users/fred/Documents/...) assumes a specific local environment and may lead to errors or unintended file access attempts on different hosting environments.

literary-master