copy-editing
Pass
Audited by Gen Agent Trust Hub on Sep 14, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill executes local Python scripts (
scripts/readability_scorer.pyandscripts/ai_content_detector.py) to provide automated feedback on marketing copy. These scripts compute standard readability metrics (Flesch Reading Ease, Flesch-Kincaid) and analyze text variance to detect potential AI generation. The scripts are entirely self-contained and do not perform network requests or access sensitive system directories.\n- [INDIRECT_PROMPT_INJECTION]: The skill possesses a data ingestion surface as it is designed to analyze user-provided marketing drafts. However, the risk is negligible due to the restricted environment.\n - Ingestion points: User-supplied marketing copy provided via CLI or file input (
draft.md).\n - Boundary markers: None explicitly defined in the instructional text.\n
- Capability inventory: Execution of local Python scripts for text processing; no network, file-write, or privilege escalation capabilities identified.\n
- Sanitization: The associated scripts utilize regex for word and sentence tokenization. While they do not filter for prompt injection, the impact is limited to statistical output, which does not allow for cross-session instruction execution or data exfiltration.
Audit Metadata