dp-dth-key-setup
Pass
Audited by Gen Agent Trust Hub on Oct 8, 2026
Risk Level: SAFECREDENTIALS_UNSAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [CREDENTIALS_UNSAFE]: The skill is designed to extract
modelPlatformApiKeycredentials from the local storage (LevelDB) of the DeepWorks desktop application. It reads sensitive data from files located in the user's%APPDATA%\com.deepexi.deepworksdirectory and migrates them to the agent's localauth.jsonconfiguration file. - [INDIRECT_PROMPT_INJECTION]: The skill ingests data from an external source (local application storage) which could be subject to poisoning.
- Ingestion points: Electron localStorage LevelDB files (
.ldband.log) located in%APPDATA%\com.deepexi.deepworksand related directories. - Boundary markers: No boundary markers or delimiters are employed when processing the ingested storage data or writing to configurations.
- Capability inventory: The script performs file-read operations on third-party application data and file-write operations on the agent's internal
auth.json,models.json, andsettings.jsonfiles. - Sanitization: The script uses regular expression matching to extract credentials and metadata but does not perform validation, escaping, or sanitization on the extracted values before writing them to the destination configuration files.
Audit Metadata