agent-task-mapping

Pass

Audited by Gen Agent Trust Hub on Feb 26, 2026

Risk Level: SAFE
Full Analysis
  • [SAFE]: No security issues detected. The skill's primary function is task delegation and agent mapping, which is implemented using standard markdown documentation and benign scripts.
  • [COMMAND_EXECUTION]: The skill includes a local Node.js script scripts/agent-discovery.js. Analysis of the source code confirms it is a safe utility that performs local file system reads (using fs.readFileSync and fs.readdirSync) to extract metadata from .agent.md files. It does not execute shell commands, perform network requests, or modify system files.
  • [PROMPT_INJECTION]: No evidence of prompt injection or bypass instructions was found in the skill definitions or subagent examples. The use of runSubagent is consistent with standard agent-to-agent communication patterns.
  • [DATA_EXFILTRATION]: No network-enabled commands or sensitive data access patterns were identified. The hardcoded path C:/Users/LOQ/.copilot/skills/ in the documentation appears to be a local configuration reference and does not pose a security risk.
Audit Metadata
Risk Level
SAFE
Analyzed
Feb 26, 2026, 10:13 PM