lody-cli
Warn
Audited by Gen Agent Trust Hub on Nov 20, 2026
Risk Level: MEDIUMDYNAMIC_EXECUTIONCOMMAND_EXECUTIONPERSISTENCEDATA_EXFILTRATIONINDIRECT_PROMPT_INJECTIONCREDENTIALS_UNSAFEEXTERNAL_DOWNLOADS
Full Analysis
- [DYNAMIC_EXECUTION]: The
lody automation register --file <check.js>command enables the registration and background execution of arbitrary JavaScript code provided by the agent to monitor conditions. - [COMMAND_EXECUTION]: The
mcp add --commandandmcp setsubcommands allow for the registration and execution of arbitrary system binaries or scripts as MCP servers on the host machine. - [PERSISTENCE]: The skill facilitates long-term persistence through the
lody schedule createcommand, which sets up recurring tasks (cron), andlody daemon install --autostart, which configures the Lody runtime as a persistent system service. - [DATA_EXFILTRATION]: The
lody workspace exportcommand generates plaintext bundles containing all conversations and attachments. Additionally,lody agent-config show --show-secretsprovides a mechanism to reveal sensitive environment variables and credentials stored in agent configurations. - [INDIRECT_PROMPT_INJECTION]: The skill acts as an orchestrator for other agents and processes untrusted data (such as external repository content via
--repoor file attachments) while maintaining high-privilege capabilities. - Ingestion points: External repository data, user prompts, and file attachments passed to
session createorsession chatcommands. - Boundary markers: Not explicitly enforced by the CLI tool; relies on the underlying agent's own constraints.
- Capability inventory: Arbitrary background JS execution, system command registration, persistence creation, and sensitive data export.
- Sanitization: Input strings and file contents are passed to the runtime without prior validation or sanitization at the CLI layer.
- [CREDENTIALS_UNSAFE]: The skill manages authentication via CLI API keys and machine connection tokens using the
lody login --auth <key>command, which can lead to credential exposure if the agent is instructed to leak or improperly handle these tokens. - [EXTERNAL_DOWNLOADS]: The skill instructions utilize
npx lodyandnpm i -g lodyto fetch and run the Lody CLI tool from the public npm registry.
Audit Metadata