The Agent Skills Directory

[COMMAND_EXECUTION]: Automatic bypass of interactive confirmation prompts. The code in src/onchainos.rs includes a --force flag when calling the wallet's contract-call function, which prevents the platform tool from requesting a final user approval before broadcasting transactions.\n- [COMMAND_EXECUTION]: Misleading CLI documentation in SKILL.md. The documentation claims write operations require a --confirm flag for execution, yet the Rust binary's source code in src/main.rs shows no implementation of such a parameter, leading to immediate execution if the dry-run flag is omitted.\n- [COMMAND_EXECUTION]: Metadata discrepancy regarding author identity. The skill metadata attributes the author as GeoGu360, whereas official vendor context identifies the author as MigOKG, suggesting potential misattribution.\n- [EXTERNAL_DOWNLOADS]: Fetches configuration and market data from a public RPC service. The skill connects to https://bsc-rpc.publicnode.com to interact with the BNB Smart Chain protocol.\n- [PROMPT_INJECTION]: Indirect prompt injection surface via on-chain data.\n
Ingestion points: Token symbols and account balances are retrieved from the blockchain via RPC calls in src/rpc.rs.\n
Boundary markers: The skill includes a 'Security Notice' in SKILL.md advising that data from the CLI be treated as untrusted.\n
Capability inventory: The skill can initiate financial transactions using the onchainos CLI wrapper.\n
Sanitization: The skill formats output as structured JSON but lacks specific filtering for potentially malicious strings embedded in blockchain metadata.

segment-finance