The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill implements a persistent state system for inventory and balances using read_file and write_file.
Ingestion points: Untrusted data enters the agent context through file reads of character state, transaction logs, and user-supplied inputs (item, seller, buyer) during BUY, SELL, and TRADE operations.
Boundary markers: The instructions do not define boundary markers or clear delimiters to differentiate between economic data and instructions.
Capability inventory: The skill uses read_file, write_file, and search_replace, allowing it to modify files based on processed data.
Sanitization: There is no evidence of sanitization or validation of the data read from external files before it is processed by the agent.
[METADATA_POISONING]: The skill includes a file named skill-snitch-report.md which contains a pre-authored security audit and a "Verdict". This is a form of metadata designed to influence or mimic security assessments, which can be misleading if taken as authoritative.

economy