The Agent Skills Directory

[CREDENTIALS_UNSAFE]: The skill documentation provides instructions for managing wallet credentials, including storing passphrases in ~/.injectived/keystore_password.txt and using the unsafe-export-eth-key command to handle raw private keys. Although the skill suggests best practices such as restrictive file permissions and avoiding logs, the management of raw cryptographic secrets is a high-risk activity.
[COMMAND_EXECUTION]: The skill includes a utility script scripts/map_injectived_cli.py that executes the injectived binary locally using subprocess.run. It recursively explores subcommands by capturing the output of help flags, which relies on the integrity of the binary being executed.
[EXTERNAL_DOWNLOADS]: The workflow recommends the installation of the injective-core package from NPM. This is a vendor-owned package from InjectiveLabs and is documented as a necessary component for the skill's primary function.
[PROMPT_INJECTION]: The command mapping script parses text output from the injectived CLI to generate documentation. This mechanism constitutes an indirect prompt injection surface where a malicious or compromised binary could produce output designed to influence the agent's behavior or poison the resulting reference material.

injective-cli