skills/bighardperson/computer-science-skills-collection/openclaw-assets-to-workbuddy/Gen Agent Trust Hub
openclaw-assets-to-workbuddy
Pass
Audited by Gen Agent Trust Hub on Oct 11, 2026
Risk Level: SAFECREDENTIALS_UNSAFEPERSISTENCEINDIRECT_PROMPT_INJECTIONDATA_EXFILTRATION
Full Analysis
- [CREDENTIALS_UNSAFE]: The skill reads and migrates sensitive authentication data, including OAuth tokens, API keys (Feishu, Slack, QQ, DingTalk), and bot secrets from
~/.openclaw/credentials/to WorkBuddy configuration files likesettings.json. - [PERSISTENCE]: The skill intentionally migrates and creates persistence mechanisms by converting OpenClaw cron job definitions (
jobs.json) into WorkBuddy automations stored in~/.workbuddy/automations/. - [INDIRECT_PROMPT_INJECTION]: The skill merges instructions from external identity files (
SOUL.md,IDENTITY.md,USER.md) which define agent persona and behavior. This presents an attack surface where malicious instructions in the source files could influence the agent's logic after migration. - Ingestion points: Files located at
~/.openclaw/workspace/SOUL.md,IDENTITY.md, andUSER.md. - Boundary markers: The instructions lack specific delimiters or escaping when merging these files, relying on the agent to decide how to "absorb" the content.
- Capability inventory: The skill has read/write access to the user's home directory (
~/.workbuddy) and IDE configuration paths. - Sanitization: No explicit sanitization or validation of the content within the identity files is performed during the merge process.
- [DATA_EXFILTRATION]: The skill accesses and processes a large volume of sensitive user data from the home directory, including identity documents, memory logs, and bot connection configurations, as part of the migration process.
Audit Metadata