agent-adaptive-coordinator

Pass

Audited by Gen Agent Trust Hub on Mar 1, 2026

Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill implements automated coordination using a suite of framework-specific commands (e.g., mcp__claude-flow__swarm_init, mcp__claude-flow__neural_train, mcp__claude-flow__memory_usage) within its pre and post hooks. These operations are aligned with the skill's primary purpose and use vendor-provided tools.\n- [PROMPT_INJECTION]: An indirect prompt injection surface is present in the task analysis logic.\n
  • Ingestion points: The skill ingests external task data via the $TASK environment variable in the pre-hook.\n
  • Boundary markers: The $TASK variable is interpolated directly into shell strings and JSON metadata without delimiters or instructions to ignore embedded instructions.\n
  • Capability inventory: The skill has permissions to initialize agent swarms, train neural models, and modify persistent memory.\n
  • Sanitization: No evidence of input validation or escaping was found for the $TASK variable before its use in execution hooks.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 1, 2026, 04:32 PM