Council

================================================================================

🟡 VERDICT: MEDIUM

This skill is rated MEDIUM due to the presence of unverified local dependencies. The skill's functionality relies on external TypeScript files (./lib/council-utils, ./lib/algorithm-utils, ./lib/observability) that are imported but whose content is not provided for security analysis. This prevents a complete audit of the skill's behavior, as these unverified files could potentially contain malicious code, including arbitrary command execution or data exfiltration.

Total Findings: 2

🟡 MEDIUM Findings: • Unverifiable Local Dependencies

• Line 206 (workflows/Convenecouncil.md): The skill imports ./lib/algorithm-utils and ./lib/observability but their content is not provided. This prevents a full security audit of the skill's behavior, as these files could contain unverified or malicious code. The sendEventToObservability call, in particular, could potentially exfiltrate data if implemented maliciously within the unverified observability.ts file. • Unverifiable Local Dependencies
• Line 20 (SKILL.md): The skill references import { conveneCouncil } from './lib/council-utils'; which is a local import whose content is not provided for analysis. This prevents a full security audit of the skill's behavior.

🔵 LOW Findings: • Indirect Prompt Injection Risk

• Line 14 (workflows/Convenecouncil.md): The skill is designed to process user-provided input (topic, context, constraints) which is then inserted into sub-agent prompts. This makes the skill inherently susceptible to indirect prompt injection, where a malicious user could craft input to manipulate the sub-agents' behavior. This is an inherent risk of the skill's design, not a direct malicious pattern within the skill's code.

================================================================================