Security risk report for @forgecat/contains-studio_agents_engineering v0.1.2
Source Integrity
Low
Profile is sourced from a legitimate public repository (contains-studio/agents on GitHub) with clear attribution and version tracking.
No supply-chain indicators of tampering; metadata (commit hash, license field, platform compatibility) is transparent and verifiable.
Agent Intent
Low
Content describes legitimate engineering roles and best practices (API design, testing, DevOps patterns, mobile development) without instructing the AI to perform malicious actions.
No prompt injection, system prompt leakage, or guidance poisoning detected; the instructions are role-descriptive and task-focused, not manipulative.
Recommendations for well-known frameworks (React, Next.js, Kubernetes, Terraform) and standard security practices (OWASP, RBAC, encryption) are legitimate and not deceptive.
Details
Evidence
ai-engineer.md: 'Design effective prompts for consistent outputs' — describes LLM best practices, not instruction to exfiltrate or hide behavior.
backend-architect.md: 'Implementing proper authentication (JWT, OAuth2)' and 'Validating and sanitizing all inputs' — standard security guidance.
test-writer-fixer.md: 'Never weakening tests just to make them pass' — reinforces code quality, not malicious intent.
Permissions
PRM-000Medium
Several agents declare Bash tool access (ai-engineer, backend-architect, devops-automator, frontend-developer, mobile-app-builder, rapid-prototyper), which is a high-risk shell execution capability.
The Bash tool grants arbitrary command execution on the host system; while plausible for DevOps/build tasks, it is broad authority that could be misused if the agent is manipulated.
The devops-automator agent's combination of Bash + Write + MultiEdit is particularly broad given its stated purpose of managing infrastructure and deployments, though it aligns with the described function.
ForgeCat