heating the forge...

@forgecat/msitarzewski_agency-agents_agency-engineering — Security Risk Report | ForgeCat

msitarzewski_agency-agents_agency-engineering

Security risk report for @forgecat/msitarzewski_agency-agents_agency-engineering v0.0.11

Low

Static scanner assigned safe level; profile is sourced from a public GitHub repository (msitarzewski/agency-agents) with clear attribution and MIT license, reducing supply-chain risk.
Version pinning (0.0.11, commit 9c31d86) and documented original source provide traceability; no evidence of typosquatted dependencies or hidden package substitution.

Low

Content describes legitimate engineering agent personalities with domain expertise (data remediation, ML, backend architecture, etc.); no instructions to ignore system prompts, read credentials, or exfiltrate data.
AI Data Remediation Engineer agent emphasizes safety guardrails ('Rule 1: AI Generates Logic, Not Data', 'Rule 2: PII Never Leaves the Perimeter', 'Rule 3: Validate the Lambda Before Execution') — these are defensive, not malicious.
Autonomous Optimization Architect discusses circuit breakers and cost guardrails to prevent runaway API calls — protective guidance, not poisoning; no instruction to weaken security or install backdoors.

Details

Findings

Invisible characters detected
agents/engineering-autonomous-optimization-architect.md · SKILL-011
Invisible characters detected
agents/engineering-backend-architect.md · SKILL-011
Invisible characters detected
agents/engineering-code-reviewer.md · SKILL-011
Invisible characters detected
agents/engineering-database-optimizer.md · SKILL-011
Invisible characters detected
agents/engineering-devops-automator.md · SKILL-011
Invisible characters detected
agents/engineering-frontend-developer.md · SKILL-011
Invisible characters detected
agents/engineering-mobile-app-builder.md · SKILL-011
Invisible characters detected
agents/engineering-rapid-prototyper.md · SKILL-011
Invisible characters detected
agents/engineering-senior-developer.md · SKILL-011
Invisible characters detected
agents/engineering-software-architect.md · SKILL-011
Invisible characters detected
agents/engineering-solidity-smart-contract-engineer.md · SKILL-011

Low

Profile declares no tools or MCP servers; agent personalities are markdown instructions only, with no executable permissions or file/shell access requested.
Descriptions reference legitimate frameworks (TensorFlow, PyTorch, Spark, dbt, PostgreSQL) and cloud platforms, but these are recommendations for the human developer to use, not permissions granted to the agent.
Code examples (PySpark, SQL, TypeScript) are illustrative templates, not instructions to execute arbitrary commands; no alwaysApply rules or glob patterns requesting broad filesystem/shell authority.

Details

Evidence

Profile metadata: 'Dependencies: None' and 'Declared tools: (none declared)'
AI Engineer: 'ML Frameworks: TensorFlow, PyTorch, Scikit-learn...' — listed as capabilities, not as installed tools.
Database Optimizer: SQL examples are educational, not executable directives.

Low

No MCP servers are defined in the profile; the artifact is purely markdown agent personalities without server binaries or network definitions.
No hidden instructions in tool descriptions, no arbitrary binary execution, and no unrestricted network/filesystem access requested.
Profile is self-contained documentation; integration with external tools (Ollama, cloud APIs) is described as guidance for the developer, not as MCP server configurations.

Details

Evidence

Profile metadata: 'MCP servers (untrusted data): (none)'
All agent definitions are markdown with no server configuration blocks or binary references.