impeccable
Pass
Audited by Gen Agent Trust Hub on Nov 3, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONREMOTE_CODE_EXECUTIONPROMPT_INJECTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill manages a platform-specific design engine binary. If not present locally, it is downloaded from a verified GitHub release repository (github.com/pbakaus/impeccable).
- [REMOTE_CODE_EXECUTION]: The skill executes a downloaded binary to perform design audits. This execution is protected by a verification process that checks the binary's SHA-256 hash against a local manifest (scripts/engine-checksums.txt) before each run. The source is a well-known service (GitHub) and is associated with the skill's developer.
- [COMMAND_EXECUTION]: The skill uses a launcher script to execute various design-related tasks, such as starting a local development server for visual audits, capturing screenshots via system tools, and managing project configuration files.
- [PROMPT_INJECTION]: Heuristic markers for concealment were identified in reference/craft-floor.md and reference/craft.md. These directives concern the suppression of internal checklists and deprecated command aliases in agent responses to improve user experience, and do not represent attempts to bypass safety guardrails.
- [INDIRECT_PROMPT_INJECTION]: The skill represents a vulnerability surface as it ingests and processes external markup and design files (Ingestion points: source code, CSS, HTML) while possessing significant file-writing and execution capabilities. However, it operates in a developer context with user-driven commands and includes instructions for least-privilege operations.
- [SAFE]: The skill follows professional standards for AI extensions, providing explicit documentation for all automated hooks, clear instructions for sensitive operations like browser debugging, and strict cryptographic validation of its own dependencies.
Audit Metadata