The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest user-supplied topic content and process it through multiple analysis modules. The resulting reports are automatically persisted to the local filesystem under the 01-内容生产/待深化的选题/ directory. This creates a vulnerability surface where a malicious payload in the user input could potentially influence the agent's behavior or be written to the local environment.
Ingestion points: User-provided "topic content" ([选题内容]) and input parameters processed by the analysis pipeline.
Boundary markers: The instructions lack explicit boundary markers (like XML tags or specific delimiters) to separate user input from the skill's own logic, increasing the risk of the agent following instructions embedded within the user's topic.
Capability inventory: The skill uses file-writing capabilities to save analysis reports and file-reading capabilities to access other local files (e.g., li-writer skill files and internal knowledge bases).
Sanitization: There is no mention of input validation, escaping, or sanitization before the content is used in prompts or written to disk.

li-topic