lov-article-creator
Pass
Audited by Gen Agent Trust Hub on Sep 29, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest external research materials, web excerpts, and partner articles for processing. It implements several robust security and fidelity measures:
- Ingestion points: Untrusted data enters the agent context during Step 1 ('Build the truth ledger') from source files and project notes.
- Boundary markers: For article reposting, the skill enforces a 'frozen zone' using
data-repost-source="true"and useswriting-review.jsonto track semantic review status. - Capability inventory: The skill executes local Python scripts for packaging and validation (
scripts/), calls an externallov-illustratetool, and writes to the local filesystem. - Sanitization: It uses
scripts/audit_repost.pyandscripts/validate_article_package.pyto ensure that source text has not been modified using SHA-256 hashing and text normalization. - [COMMAND_EXECUTION]: The skill instructs the agent to execute several local Python scripts provided in the
scripts/directory to handle deterministic logic such as image processing (Pillow), YAML parsing, and file integrity checks. These scripts are strictly scoped to the article packaging task.
Audit Metadata