viral-title

Pass

Audited by Gen Agent Trust Hub on Jun 15, 2026

Risk Level: SAFECOMMAND_EXECUTION
Full Analysis
  • [SAFE]: The skill provides a secure and structured environment for its stated task. It relies on internal reference files and local scripts, with no external network dependencies or unsafe data processing identified.
  • [COMMAND_EXECUTION]: The skill executes local Python scripts to perform logging, retrieval, and analysis tasks. Scripts such as "scripts/analyze_feedback.py" and "scripts/retrieve_title_examples.py" use Python's standard library for file operations and do not employ unsafe functions like "eval" or "exec".
  • [PROMPT_INJECTION]: The skill features a feedback loop which presents an indirect prompt injection surface. This was evaluated with the following findings: 1. Ingestion points: User feedback and edited titles enter the context via "SKILL.md" and are stored in "references/evolution/feedback.jsonl". 2. Boundary markers: Absent. 3. Capability inventory: Local file-write operations and log aggregation are performed by "scripts/log_feedback.py" and "scripts/analyze_feedback.py". 4. Sanitization: Present; data is stored using structured JSON serialization. The risk is significantly mitigated by core rules in "meta/RULES.md" that prohibit automatic methodology updates without explicit user confirmation.
Audit Metadata
Risk Level
SAFE
Analyzed
Jun 15, 2026, 06:32 AM
Security Audit — agent-trust-hub — viral-title