kelly-wechat-crm
Fail
Audited by Gen Agent Trust Hub on Aug 26, 2026
Risk Level: HIGHEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONDATA_EXFILTRATION
Full Analysis
- [UNVERIFIABLE_DEPENDENCIES_AND_REMOTE_CODE_EXECUTION]: The synchronization script 'scripts/sync_wechat.mjs' includes instructions and error handling that direct the user to install a binary from an untrusted external domain ('wechat-cli.com') using the dangerous 'curl -fsSL ... | sh' pattern.
- [PRIVILEGE_ESCALATION]: In 'scripts/sync_wechat.mjs', the code provides error messages explicitly suggesting that the operator use 'sudo' to run initialization commands, which facilitates potential unauthorized administrative access to the host machine.
- [DATA_EXPOSURE_AND_EXFILTRATION]: The skill's primary function involves reading sensitive personal data from a local WeChat installation (contacts and message session summaries) and transmitting it to a remote Busabase instance via the 'busabase-sdk'.
- [COMMAND_EXECUTION]: The script 'scripts/sync_wechat.mjs' utilizes 'execFileSync' to execute the local 'wechat-cli-rs' binary, which introduces a vector for local command execution.
- [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from external sources (WeChat messages and contact names). * Ingestion points: 'scripts/sync_wechat.mjs' (processes local binary JSON output). * Boundary markers: Absent. * Capability inventory: Uses 'execFileSync' and performs network writes via 'busabase-sdk'. * Sanitization: No sanitization or escaping of message content was found before the data is integrated into the agent context.
Recommendations
- AI detected serious security threats
Audit Metadata