kill-the-wall-of-text
Pass
Audited by Gen Agent Trust Hub on Oct 13, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [COMMAND_EXECUTION]: The skill includes a Python validation script (
check.py) inreferences/pre-publish-check.md. The instructions direct the agent or user to execute this script locally to verify that the generated HTML output complies with the Simplified Technical English (STE) and fidelity rules. - [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process arbitrary user-provided content, such as meeting notes or reports, which presents a potential injection surface.
- Ingestion points: Untrusted source text provided by the user for processing.
- Boundary markers: The prompt interpolation for source content does not utilize explicit boundary markers or delimiters to isolate user data from agent instructions.
- Capability inventory: The skill can write local HTML files and suggests the execution of a shell script for validation.
- Sanitization: While no technical sanitization is mentioned, the skill enforces a strict 'Simplified Technical English' output format which naturally limits the effectiveness of embedded malicious instructions.
- [EXTERNAL_DOWNLOADS]: The provided HTML skeleton in
references/page-skeleton.mdallows the artifact to fetch assets like fonts or chart libraries from well-known public CDNs if needed for the document's presentation.
Audit Metadata