phaser4-game-workflow-control
Pass
Audited by Gen Agent Trust Hub on Oct 23, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill interacts with the local Git repository to manage state and verify candidate changes.
- Evidence:
scripts/execution-unit-control.mjsusesexecFileSyncto rungit diffandgit ls-filesvia a providediowrapper. - Security Note: Commands use argument arrays instead of shell strings, which is a safe practice that prevents typical command injection vulnerabilities.
- [INDIRECT_PROMPT_INJECTION]: The skill ingests complex structured data from external files (Work Items, Implementation Packages) which influences its logic and state transitions.
- Ingestion points:
scripts/runtime/io.mjsreads various JSON files from the project directory. - Boundary markers: Strict JSON Schema enforcement (
references/*.schema.json) is used for all input types to ensure data integrity and prevent schema confusion. - Capability inventory: The skill is capable of reading/writing project files and executing git commands within the project scope.
- Sanitization:
scripts/visual-contract-core.mjsandscripts/visual-package-paths.mjsimplement robust path normalization (normalizeContractPath) that blocks absolute paths and directory traversal attempts, ensuring operations remain within the project boundary.
Audit Metadata