agent-package-manager
Pass
Audited by Gen Agent Trust Hub on Apr 5, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill instructs the agent to operate the
apmCLI, including subcommands likeapm run, which executes project-defined scripts, andapm runtime setup, which manages local environments. - [EXTERNAL_DOWNLOADS]: The troubleshooting documentation provides instructions to download the APM toolchain from Microsoft's official short-link service (
aka.ms). - [REMOTE_CODE_EXECUTION]: The installation guide includes patterns for piping remote scripts directly to shell interpreters (
curl | shandirm | iex) from Microsoft's infrastructure. - [PROMPT_INJECTION]: The skill has an indirect prompt injection surface as it is designed to fetch, process, and integrate external agent packages, skills, and prompts from remote repositories.
- Ingestion points: External repository content (GitHub, GitLab, etc.) fetched via
apm installand manifest files (apm.yml). - Boundary markers: No explicit delimiters or instructions to ignore embedded commands are specified for the integrated content.
- Capability inventory: The toolchain possesses file system write access, network capabilities for dependency resolution, and the ability to execute project scripts.
- Sanitization: There is no evidence of sanitization or safety validation for natural language instructions retrieved from remote sources before they are compiled into the agent's active context.
Audit Metadata