The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill downloads the necessary command-line tools and installer scripts from vendor-owned repositories on GitHub. The installation process in SKILL.md is secured by SHA256 checksum verification, ensuring that the downloaded code has not been tampered with before it is executed on the system.
[COMMAND_EXECUTION]: The strategy utilizes the onchainos CLI to perform wallet operations, fetch market prices, and scan token rankings. These operations are conducted through subprocess calls from the Python engine to the local binary, which is a standard method for integrating specialized tools into AI agent skills.
[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from external sources, including token metadata and market leaderboards. • Ingestion points: External data enters the system in ranking_sniper.py through CLI commands such as 'token trending' and 'advanced-info'. • Boundary markers: No explicit text delimiters are used for the external data, but the content is treated as structured JSON data rather than natural language instructions. • Capability inventory: The skill has the capability to execute on-chain swaps, write to local state files, and host a local web dashboard. • Sanitization: The skill performs extensive validation on the ingested data, filtering tokens through 25 distinct safety checks and numeric thresholds before any trade actions are taken.
[SAFE]: The skill follows security best practices for financial agents, including a default 'paper trading' mode, requirement for typed user confirmation to enable live trading, and the use of a Trusted Execution Environment (TEE) for secure transaction signing.

top-rank-tokens-sniper