The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill facilitates the processing of data from the NEAR blockchain, which is an untrusted external source that could contain embedded instructions.
Ingestion points: Data is ingested through contract view methods such as near.view() and React hooks like useView() as seen in SKILL.md and references/react.md.
Capability inventory: The skill has the capability to perform sensitive actions including sending tokens, calling contract methods, and deploying code via near.call(), near.send(), and the transaction() builder (SKILL.md, references/transactions.md).
Boundary markers: The provided implementation examples do not include explicit boundary markers or instructions for the agent to ignore content within retrieved blockchain data.
Sanitization: No sanitization or validation logic is demonstrated for external data before it is processed by the agent or used in transactions.
[CREDENTIALS_UNSAFE]: The documentation references accessing a sensitive filesystem path commonly used for storing private keys.
Examples for the FileKeyStore configuration point to the ~/.near-credentials directory, which is the default location for NEAR Protocol secret keys (SKILL.md, references/keys-and-testing.md). While this is standard for the protocol's developer tools, it involves the handling of sensitive credential files.
[EXTERNAL_DOWNLOADS]: The skill documentation includes instructions to download and install various Node.js packages.
Dependencies include the library itself (near-kit), its React bindings (@near-kit/react), and standard ecosystem components like @near-wallet-selector/core and @hot-labs/near-connect (references/react.md, references/wallets.md).
[COMMAND_EXECUTION]: The instructions involve executing shell commands for package installation.
The documentation provides npm install commands to set up the required environment and library dependencies (references/react.md, references/wallets.md).

near-kit