dby-banned-words
Pass
Audited by Gen Agent Trust Hub on Sep 9, 2026
Risk Level: SAFEDATA_EXFILTRATIONCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [DATA_EXFILTRATION]: The skill transmits user-provided text to the external API at
https://doubaoya.com/api/apis/tool/check-banned-words/call. This is the documented core functionality of the skill and is required for content analysis. - [COMMAND_EXECUTION]: The skill executes a local script
scripts/check_multi.pyusing Python 3. The script acts as a wrapper for network requests and does not perform any dangerous file system or system-level operations. - [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process untrusted text, which introduces a potential attack surface for indirect prompt injection.
- Ingestion points: User-provided copy is passed as a command-line argument to
scripts/check_multi.pyand subsequently sent to the remote API. - Boundary markers: The instructions lack formal delimiters (like XML tags or specific delimiters) for the untrusted content when instructions are given to the agent for rewriting the copy.
- Capability inventory: The skill can execute local Python scripts and perform network operations to the tool provider's domain.
- Sanitization: The Python script uses
json.dumps()to serialize the content, which provides standard escaping for characters relevant to the JSON payload.
Audit Metadata