The Agent Skills Directory

[SAFE]: No critical or high-risk security threats, such as credential theft, privilege escalation, or malicious persistence, were detected within the skill's instructions or scripts.
[EXTERNAL_DOWNLOADS]: The skill downloads knowledge base content in Markdown format from a documented GitHub repository (dontbesilent2025/dbskill). The scripts/tc_knowledge.py script fetches this content for local indexing and search; the downloaded data is parsed as text and is not executed as code.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process data from external repositories and user-connected platforms (e.g., Feishu/Lark, Notion).
Ingestion points: Content is retrieved from remote URLs and local directories into the agent's context via the tc_knowledge.py script.
Boundary markers: The SKILL.md provides explicit instructions to isolate historical data and external perspectives, preventing them from influencing core decision-making unless explicitly requested by the user.
Capability inventory: The skill performs file I/O and network GET requests; it does not utilize dynamic code execution functions like eval or exec on ingested content.
Sanitization: External data is parsed into structured formats using regular expressions and JSON/JSONL loaders, ensuring content is treated as data rather than instructions.
[COMMAND_EXECUTION]: The skill utilizes a local Python script (scripts/tc_knowledge.py) for management tasks. These operations are restricted to standard data processing, searching, and maintaining a local cache in the user's home directory (~/.tc), which follows best practices for tool-assisted knowledge management.

tc-knowledge