placeholder-token-networks
Pass
Audited by Gen Agent Trust Hub on Aug 31, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONNO_CODE
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes external information such as protocol research, on-chain data, and market context, which serves as a potential vector for indirect prompt injection.
- Ingestion points: The 'Inputs Needed' section in
SKILL.mdspecifies that the agent should ingest asset context, research data, and adoption data. - Boundary markers: The skill includes guardrails that instruct the agent to separate facts from assumptions and follow a rigid output format, which acts as a soft boundary marker.
- Capability inventory: There are no high-risk capabilities such as file system writes, network exfiltration tools, or command execution defined in the skill.
- Sanitization: No specific input sanitization or delimiter strategies are defined for the external data context.
- [NO_CODE]: The skill consists entirely of Markdown instructions and configuration files (
SKILL.mdandinvest.md) without any scripts, binaries, or executable code, significantly reducing the attack surface.
Audit Metadata