souldinals
Pass
Audited by Gen Agent Trust Hub on Oct 1, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill fetches inscription content, metadata, and transaction information from the Unisat Ordinals API (https://open-api.unisat.io) and refers users to mempool.space for transaction monitoring.
- [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted Markdown content from the Bitcoin blockchain which can influence agent behavior.
- Ingestion points: Raw inscription content is fetched from the Unisat API within the
load-soulanddisplay-soulcommands insouldinals/souldinals.ts. - Boundary markers: Ingested data is returned to the agent without delimiters or instructions to ignore potential instructions embedded in the blockchain data.
- Capability inventory: The
AGENT.mdinstructions explicitly suggest using the loaded content to "reload agent identity context." Additionally, the skill has capabilities to sign and broadcast Bitcoin transactions, potentially allowing an attacker to trigger unauthorized transfers if the agent is compromised by injected instructions. - Sanitization: The
parseSoulTraitslogic extracts text directly from Markdown headers and lists without sanitizing for prompt injection patterns or malicious instructions.
Audit Metadata