foundry-security-spec
Pass
Audited by Gen Agent Trust Hub on Sep 18, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill instructs users to install a specification toolkit directly from the official GitHub organization via its public package registry using the command
npm install -g @github/spec-kit.\n- [INDIRECT_PROMPT_INJECTION]: The multi-agent architecture described is inherently designed to ingest and analyze untrusted source code from external repositories for vulnerability discovery, creating a potential surface for indirect prompt injection.\n - Ingestion points: Agent roles such as the
Orchestrator,Detector, andExploreringest repository contents and code snippets (unit.content) for processing.\n - Boundary markers: The specification enforces a "Foundry Constitution" with the core principle of "No unsupervised execution," requiring explicit human confirmation before any finding is accepted.\n
- Capability inventory: The defined agents are designed to communicate with external issue trackers (Jira, GitLab) and LLM providers via configurable gateways.\n
- Sanitization: The implementation guidelines mandate that all novel security hypotheses and validation steps be performed within isolated Docker containers with strict network isolation to prevent impact on the execution environment.
Audit Metadata