skills/mukul975/anthropic-cybersecurity-skills/performing-web-application-vulnerability-triage/Gen Agent Trust Hub
performing-web-application-vulnerability-triage
Pass
Audited by Gen Agent Trust Hub on Sep 14, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process scan results (JSON files) from DAST and SAST tools, which are external and potentially attacker-influenced sources.
- Ingestion points: The
scripts/agent.pyscript ingests external findings via theingest_json_reportmethod, loading data directly from files provided as command-line arguments. - Boundary markers: The instructions and script do not implement explicit boundary markers or "ignore" instructions to prevent the agent from following commands that might be embedded in the scanner findings.
- Capability inventory: The skill includes Python scripts for data processing and prioritization. The
SKILL.mdfile also describes workflows that involve pushing findings to external management tools like DefectDojo or Jira. - Sanitization: The script implements minimal sanitization by truncating descriptions to 500 characters, but it lacks robust validation or escaping of fields such as
title,url, andparameterbefore they are included in the triage report output.
Audit Metadata