threat-modeling-expert — agentic threat model

Not certain from the listing — likely relies on third-party foundation models. Threats include prompt injection to bypass security analysis or force the model to generate biased, weak, or intentionally flawed threat models.

Not certain from the listing — likely ingests architecture designs, codebases, or documentation. Threats include data exfiltration of highly sensitive system designs, or knowledge-base poisoning to make the agent ignore specific vulnerabilities.

Not certain from the listing — as an 'Agent Skill', it integrates into an orchestration framework. Threats include insecure tool integration if it calls external parsers for architecture diagrams, or framework vulnerabilities allowing execution of malicious payloads embedded in design docs.

Not certain from the listing — hosted as an open-source skill. Threats include container compromise if run in an unsandboxed environment, or exposure of the API hosting the skill.

Not certain from the listing — no built-in evaluation or guardrails mentioned. Threats include blind spots where the model fails to identify critical threats, or lack of audit logs for security reviews.

Not certain from the listing — lacks explicit authentication or access controls in the description. Threats include unauthorized access to sensitive threat models and lack of regulatory compliance (e.g., GDPR/NIST) regarding design data.

Not certain from the listing — could be used in multi-agent setups for automated dev pipelines. Threats include rogue agents querying this skill to find zero-days in a system, or cascading failures if its threat reports are automatically ingested by patching agents.