The Agent Skills Directory

[COMMAND_EXECUTION]: The skill makes extensive use of the bounded CLI tool to perform all operations, including project setup, deployment, data plane interactions, and environment management.
[EXTERNAL_DOWNLOADS]: The skill fetches the Bounded CLI binary and updates from the vendor's official release domain at get.bounded.sh.
[REMOTE_CODE_EXECUTION]: The documentation provides instructions to install or update the platform CLI using a curl | sh pattern from the vendor's HTTPS release host.
[CREDENTIALS_UNSAFE]: The skill interacts with sensitive local files in ~/.bounded/, including Solana ed25519 secret keys and web session tokens. It provides guidance on using chmod for file safety and automatically maintains .gitignore rules to prevent these secrets from being committed to version control.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted input from local project files and CLI error responses to guide the agent's workflow. It also accepts natural language prompts for app creation and editing which are sent to the platform's build agent. Ingestion points: Reads bounded.json, policy.json, and terminal output from the bounded tool. Boundary markers: The skill uses backticks for commands and structured sections to separate instructions from data. Capability inventory: The skill allows for filesystem access, shell command execution, and network operations through the platform CLI. Sanitization: The CLI tool performs schema validation and policy verification via the bounded verify command.
[DYNAMIC_EXECUTION]: The skill includes a Python script to parse and display metadata from a local JSON session file.

bounded-deploy