gimble-release
Pass
Audited by Gen Agent Trust Hub on Sep 23, 2026
Risk Level: SAFEREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONPERSISTENCEINDIRECT_PROMPT_INJECTION
Full Analysis
- [REMOTE_CODE_EXECUTION]: The skill utilizes
vp dlxto download and install agent skills directly from thetylergannon/gimbleGitHub repository. It also performs remote installation of the application's Go binary using thego installcommand targeting the author's repository. - [COMMAND_EXECUTION]: The instructions direct the agent to execute various shell commands for the build and deployment process, including
just build,go install, andvp dlxfor environmental setup. - [PERSISTENCE]: The skill automates the installation of files into the persistent agent environment at
~/.agents/skills/and installs thegimblebinary into the user's executable path, which maintains its state across system sessions. - [INDIRECT_PROMPT_INJECTION]: The skill defines a workflow for reading and processing local documentation and requirement files, which serves as a potential surface for indirect prompt injection.
- Ingestion points: Reads
AGENTS.md,docs/definition-of-done.md, and local requirement files in the worktree. - Boundary markers: Absent; the instructions do not specify delimiters or warnings for these inputs.
- Capability inventory: Includes command execution via
justandgo, file system manipulation via installation commands, and network access for skill retrieval. - Sanitization: Absent; the agent is instructed to use outcomes and facts derived from these files to guide implementation.
Audit Metadata