rrragflow-skill

Use for RAGFlow dataset tasks: create, list, inspect, update, or delete datasets; upload, list, update, or delete documents; start or stop parsing; check par...

This skill appears to do what it says: it runs the included Python scripts against the RAGFlow API you point it at and requires only RAGFLOW_API_URL and RAGFLOW_API_KEY. Before installing, consider: (1) Only provide an API key for a RAGFlow instance you trust — the scripts will send that key to whatever URL you configure. (2) Upload commands read local files you specify (upload.py); don't upload sensitive files to a server you don't trust. (3) The SKILL.md requests the agent require explicit confirmation for deletes; the CLI requires --ids but the ultimate safety depends on the agent following that guardrail — confirm deletion prompts with the agent before proceeding. (4) If you need stronger guarantees, review upload.py and other scripts locally to confirm they only transmit intended files/fields. Overall, nothing in the package is disproportionate to its stated purpose.

Type: OpenClaw Skill Name: rrragflow-skill Version: 1.0.4 The skill bundle provides legitimate functionality for managing RAGFlow datasets and documents, but it is classified as suspicious due to the presence of high-risk capabilities that are explicitly mentioned in the analysis criteria. Specifically, the scripts `upload.py`, `update_dataset.py`, and `update_document.py` allow for arbitrary local file reading and uploading, including a feature to load JSON payloads from local paths using a '@' prefix. While these capabilities are plausibly needed for the stated purpose of document ingestion and configuration, they represent a significant attack surface for prompt injection, as an agent could be coerced into exfiltrating sensitive local files (e.g., JSON configuration files or credentials) to the user-provided `RAGFLOW_API_URL` endpoint.