devin-cli
Pass
Audited by Gen Agent Trust Hub on Sep 22, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: Fetches the Devin CLI installation script from Cognition AI's official domain (cli.devin.ai) using a standard shell-piped execution pattern for installation.
- [COMMAND_EXECUTION]: Executes the local
devinbinary via the terminal to perform coding tasks, including headless runs with auto-approval modes designed for automation. - [INDIRECT_PROMPT_INJECTION]: Ingests task instructions from external files that may contain untrusted data.
- Ingestion points: The
devin -p --prompt-fileflag inSKILL.mdloads instructions from a brief file. - Boundary markers: The skill defines a mandatory 'Prompt contract' to restrict the agent's scope, including rules for worktree boundaries and stopping criteria.
- Capability inventory: The Devin CLI possesses broad capabilities including file system access, network operations, and command execution.
- Sanitization: The skill mitigates risks by requiring the use of isolated git worktrees and recommends reviewing diffs and rerunning tests before integration.
Audit Metadata