Docker Expert
Pass
Audited by Gen Agent Trust Hub on Oct 20, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze untrusted data from the user's project environment, including Dockerfiles, Compose configuration files, and file system structures. This creates a potential vector for indirect prompt injection where malicious instructions could be embedded in the analyzed content.\n
- Ingestion points: The skill reads local project files and container metadata using find, grep, docker ps, and docker images via the SKILL.md instructions.\n
- Boundary markers: The instructions lack explicit delimiters or guidance to distinguish between the skill's internal instructions and the potentially untrusted content being analyzed.\n
- Capability inventory: The skill has the ability to execute shell commands, perform network requests (via examples containing curl), and manage Docker containers (build, run, exec).\n
- Sanitization: There is no evidence of sanitization or filtering of the content read from user-provided files before it is processed by the agent.\n- [COMMAND_EXECUTION]: The skill uses local shell commands to inspect the Docker environment and validate container builds.\n
- Evidence: The skill executes docker info, docker ps, find, docker build, and docker run to perform its diagnostic and validation tasks.
Audit Metadata