The Agent Skills Directory

[SAFE]: The skill follows secure practices by storing data in a dedicated local directory (~/.dbs) and providing a privacy mode to handle sensitive information using aliases. Analysis of the instructions and operational logic confirms no presence of malicious patterns, obfuscation, or unauthorized network activity.
[PROMPT_INJECTION]: The skill processes local Markdown files to maintain state. While this presents a potential indirect prompt injection surface if malicious content is manually added to these files, the risk is inherent to the skill's legitimate purpose of persistent knowledge management. 1. Ingestion points: Reads 我的当前状态.md, 00_使用说明.md, and AGENTS.md during initialization and project updates. 2. Boundary markers: Not explicitly defined in the file format for ingestion. 3. Capability inventory: File system operations limited to the skill's working directory. 4. Sanitization: Relies on user/AI following the defined Markdown structure.

dbs-decision