docker-extend
Pass
Audited by Gen Agent Trust Hub on Sep 15, 2026
Risk Level: SAFEDYNAMIC_CONTEXT_INJECTIONINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [DYNAMIC_CONTEXT_INJECTION]: The skill utilizes shell command execution blocks (using the
!syntax) to detect whether the user is in a development or deployment environment and to read the contents of existing Docker configuration files. These commands are restricted to state inspection and project structure detection, facilitating better context for the agent. - [INDIRECT_PROMPT_INJECTION]: The skill ingests user-provided tool names as arguments and interpolates them into a
Dockerfile. While this allows user input to influence configuration files, the risk is mitigated because the agent is responsible for parsing the input into a specific block format, and the resulting changes are restricted to local, non-committed files. - [COMMAND_EXECUTION]: The skill suggests commands for building and starting Docker containers (e.g.,
docker compose up). It explicitly avoids automatic execution of these potentially heavy or environment-altering commands, ensuring that the user reviews and manually triggers the container updates.
Audit Metadata