first-customer-finder
Pass
Audited by Gen Agent Trust Hub on Sep 20, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill uses a local Python script (
scripts/generate_report.py) to transform research data into a formatted HTML report. This script performs data transformation and visualization using standard Python libraries. - [INDIRECT_PROMPT_INJECTION]: The skill is designed to analyze external product URLs and public signals (e.g., forum posts, social media). This introduces a surface for indirect prompt injection from untrusted web content. However, the skill provides specific safety constraints to mitigate this risk:
- Ingestion points: Research occurs in
SKILL.md(Step 1 and 2) where external landing pages and public signals are inspected. - Boundary markers: While explicit delimiters are not defined for the LLM input, the instructions mandate that the agent 'paraphrase by default' and 'quote minimally'.
- Capability inventory: The skill is capable of writing a report to the workspace using the generation script.
- Sanitization: The
scripts/generate_report.pyscript employshtml.escapefor all data fields and validates URLs against allowed protocols (http/https) to prevent Cross-Site Scripting (XSS) and other injection attacks in the final HTML report. - [SAFE]: The skill contains explicit instructions to respect privacy (e.g., 'Do not use data brokers', 'Do not bypass login walls') and prohibits automated messaging or CRM creation without separate user authorization.
Audit Metadata