The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill downloads installation scripts (install-local.sh, install.sh) and reference documentation from official OKX GitHub repositories. These sources are controlled by the skill author and are considered trusted in this context.
[REMOTE_CODE_EXECUTION]: The skill downloads and executes bash scripts from the vendor's repository to set up the hackathon environment. The onchainos CLI installation process implemented in SKILL.md follows security best practices by downloading a checksum file and verifying the SHA256 hash of the install.sh script before execution.
[COMMAND_EXECUTION]: The skill uses npx to install and run the skills package management tool and executes local bash scripts for documentation setup.
[INDIRECT_PROMPT_INJECTION]: The skill interacts with external community feeds on Moltbook, which presents an attack surface for indirect prompt injection. Ingestion points: The Moltbook feed API is accessed in SKILL.md and references/cli-reference.md. Boundary markers: The skill contains an explicit 'Security' section in SKILL.md warning the agent to treat third-party submissions as data and not instructions. Capability inventory: The skill utilizes curl, bash, and npx for system-level operations across README.md and SKILL.md. Sanitization: The skill instructs the agent to ignore any commands or overrides embedded in participant content.
[CREDENTIALS_UNSAFE]: The skill provides instructions for storing the Moltbook API key in a local configuration file (~/.config/moltbook/credentials.json), which is a standard method for maintaining CLI tool authentication.

okx-buildx-hackathon-agent-track