dcg
Pass
Audited by Gen Agent Trust Hub on Nov 4, 2026
Risk Level: SAFEREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONPERSISTENCEINDIRECT_PROMPT_INJECTION
Full Analysis
- [REMOTE_CODE_EXECUTION]: The validation script
scripts/validate-dcg.shcontains a printed recommendation for acurl | bashinstallation command that downloads an install script from the author's GitHub repository. Additionally, thedcg updatecommand documented inreferences/COMMANDS.mddescribes a self-update mechanism that downloads and installs binary updates from the vendor's infrastructure. - [COMMAND_EXECUTION]: The skill facilitates interaction with the
dcgCLI binary to perform health checks (dcg doctor), dry-run command evaluations (dcg test), and rule explanations (dcg explain). - [PERSISTENCE]: Documentation in
references/CONFIG.mdandreferences/TROUBLESHOOTING.mddescribes the registration of aPreToolUsehook within the agent's configuration files (e.g.,~/.config/claude-code/settings.json) to enable intercepting shell commands before execution. - [INDIRECT_PROMPT_INJECTION]: The skill acts as a security enforcement layer to prevent accidental or malicious execution of destructive commands that may be embedded in untrusted data, using a modular rule pack system and AST-based scanning for heredocs and inline scripts.
Audit Metadata