agent-manager-skill
Warn
Audited by Gen Agent Trust Hub on Oct 29, 2026
Risk Level: MEDIUMEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill requires downloading external code from a third-party GitHub repository (
https://github.com/fractalmind-ai/agent-manager-skill.git). This source is not associated with the skill author ('pacman-clyde') and does not belong to a known trusted organization. - [COMMAND_EXECUTION]: The instructions direct the agent to execute Python scripts (
python3 agent-manager/scripts/main.py) located within the downloaded repository, creating a risk of running unverified code. - [INDIRECT_PROMPT_INJECTION]: The skill has an ingestion surface for potentially untrusted data that could influence agent behavior.
- Ingestion points: The
assigncommand inSKILL.mduses a shell heredoc to pass task descriptions to agents. - Boundary markers: There are no boundary markers or delimiters specified to distinguish between the task instructions and potentially malicious content within the input.
- Capability inventory: The skill uses
tmuxto manage local CLI agents and executes Python scripts to control them. - Sanitization: No input sanitization or validation is evident in the prompt instructions or command structure.
Audit Metadata