scaffold
Pass
Audited by Gen Agent Trust Hub on Nov 4, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process external data—specifically project plans, architecture decisions, and domain models—which are then used to generate code and trigger further agent actions.
- Ingestion points: The workflow starts by reading architecture decisions and plans (Workflow Step 1, 'Read Plan').
- Boundary markers: The instructions lack explicit requirements for using delimiters or 'ignore embedded instructions' warnings when processing these plans.
- Capability inventory: The skill has high-privilege capabilities including writing new source code files, executing shell commands for builds and tests, and spawning subagents via the
spawn_agentauthorization declared in the header. - Sanitization: There is no evidence of sanitization or strict schema validation for the plan data before it influences code generation.
- [DYNAMIC_EXECUTION]: The skill's primary function is to generate code ('scaffolding') based on templates and plan inputs. It also references the execution of scripts such as
ai-signal-scan.cjsandsync-update-blocks.pywhich are expected to be present in the environment. - [COMMAND_EXECUTION]: The skill utilizes shell commands for discovery and verification, such as
grepcommands used in the 'Activation Guards' to search for existing foundational code patterns (e.g.,grep abstract class.*Base|Base[A-Z]\w+|Abstract[A-Z]\w+), and project-specific build/test commands executed during the 'Verification Gate'.
Audit Metadata