The Agent Skills Directory

[PROMPT_INJECTION]: The skill processes untrusted external data in the form of 'demand' and 'fulfillment' text. When using the --auto mode, this content is evaluated by an LLM to determine arbitration outcomes. Malicious actors could embed instructions within these texts to manipulate the agent's decision-making process.
Ingestion points: The nla escrow:arbitrate command reads arbitrary text from the NLA protocol (file: SKILL.md).
Boundary markers: No explicit delimiters or instructions are provided to the agent to ignore embedded commands within the agreement data.
Capability inventory: The skill uses the nla tool which performs network operations to interact with blockchain contracts and LLM APIs.
Sanitization: There is no evidence of sanitization or validation of the input text before it is processed by the LLM.
[EXTERNAL_DOWNLOADS]: The skill requires the installation of the nla command-line utility from the npm registry.
[COMMAND_EXECUTION]: The skill relies on executing various nla subcommands (e.g., wallet:show, escrow:arbitrate, escrow:status) which involve network communication and local state management.

nla-arbitrate