The Agent Skills Directory

[PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface because it ingests untrusted data from external sources and provides the agent with high-privilege capabilities such as financial transactions and social media posting.\n
Ingestion points: Data enters the agent context through Twitter searches and mentions (x_search, x_get_mentions in bin/solana-wallet.ts) and the pump.fun websocket feed (screener.ts).\n
Boundary markers: The CLI wrapper and plugin tools do not implement delimiters or 'ignore instructions' warnings for external content.\n
Capability inventory: The skill allows the agent to execute shell commands via spawn, perform wallet transfers and swaps on Solana and Polygon, and post content to X (Twitter).\n
Sanitization: No specific sanitization or filtering of external input was identified before interpolation into agent prompts.\n- [CREDENTIALS_UNSAFE]: The setup script (scripts/setup-master-key.ts) prints the user's master encryption password directly to the terminal console during configuration. While intended to help users populate their environment files, this practice can lead to credential exposure if terminal logs are stored, cached, or shared.\n- [COMMAND_EXECUTION]: The skill uses child_process.spawn to manage background daemon processes for its weather arbitrage and Twitter monitoring strategies. It specifically re-spawns its own CLI entry point (bin/solana-wallet.ts) as a detached process and uses pkill to terminate existing instances.\n- [EXTERNAL_DOWNLOADS]: The skill is designed to be installed from a remote GitHub repository (https://github.com/inspi-writer001/raphael-solana). It also makes legitimate network requests to established services including Solana RPC nodes, the Polygon network, and the Twitter API.

solana-wallet