The Agent Skills Directory

Data Exposure & Exfiltration (HIGH): The skill references specific local file paths for sensitive data extraction.
Evidence: Hardcoded URI file:///Users/alice/agent-o-rama/agent-o-rama/dev/secure_wallets.json in the 'See Also' section. Accessing wallet configuration files using absolute paths to a specific user directory is a significant privacy and security risk.
Indirect Prompt Injection (HIGH): The skill creates a dangerous attack surface by combining external data ingestion with transaction execution capabilities.
Ingestion points: Reads from bib.duckdb and external markdown files like WEV_SYNTHESIS.md and SKILL_ADJUNCTIONS.md.
Boundary markers: No delimiters or safety instructions are defined for processing the content of these external files.
Capability inventory: Possesses the capability to execute blockchain transactions via just wev-transfer and integration with aptos-agent.
Sanitization: No evidence of sanitization or validation for the 'knowledge differentials' or 'epistemic transfer' data processed by the agent.
Command Execution (MEDIUM): The skill relies on the just command runner to execute various operations (wev-scan, wev-transfer, aptos-world-balances). This abstracts the actual logic into an unexamined justfile, which could contain malicious shell commands.

wev-orderless