The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill requires the use of the @t2000/cli package, which is downloaded and executed from the npm registry. This package does not originate from a known trusted organization or a well-known service.
Evidence: Metadata requires t2000 CLI (npx @t2000/cli init) in SKILL.md.
[REMOTE_CODE_EXECUTION]: The use of npx to initialize the t2000 CLI tool facilitates the direct execution of remote code from the npm registry during the setup process.
Evidence: requires: t2000 CLI (npx @t2000/cli init) in SKILL.md.
[COMMAND_EXECUTION]: The skill makes extensive use of CLI commands to interact with the Sui blockchain, manage wallet balances, and execute attacks, which involves network communication and financial asset management.
Evidence: Commands such as t2000 balance, t2000 swap, and t2000 sentinel attack are used throughout SKILL.md.
[PROMPT_INJECTION]: The skill has an attack surface for indirect prompt injection because it processes untrusted output from external AI "sentinels" that could contain malicious instructions.
Ingestion points: Data enters the agent context through the outputs of t2000 sentinel list, t2000 sentinel info, and t2000 sentinel attack in SKILL.md.
Boundary markers: No delimiters or explicit instructions to ignore embedded content are present in the command flows.
Capability inventory: The skill has the capability to execute shell commands and perform financial transactions (swaps and transfers) via the CLI in SKILL.md.
Sanitization: There is no evidence of validation or sanitization of the content returned from external sentinel responses.

t2000-sentinel