Hedera Hackathon PRD
Pass
Audited by Gen Agent Trust Hub on Feb 18, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [Indirect Prompt Injection] (LOW): The skill ingests untrusted text in Step 1 when asking users to paste bounty or track descriptions. This presents a surface for indirect prompt injection where malicious instructions could be embedded in the data. However, the potential impact is limited to the content of the generated PRD document. * Ingestion points: SKILL.md Step 1. * Boundary markers: Absent; there are no delimiters or specific instructions for the agent to ignore commands within the pasted text. * Capability inventory: The skill is limited to writing a markdown file (HACKATHON-PRD.md). * Sanitization: No input validation or sanitization is performed.
- [Data Exposure & Exfiltration] (SAFE): No credentials, sensitive file paths, or network exfiltration patterns were detected.
- [Remote Code Execution] (SAFE): The skill does not download or execute any external scripts or software packages.
Audit Metadata