The Agent Skills Directory

[COMMAND_EXECUTION]: The skill provides the agent with extensive authority to execute shell commands via the @alchemy/cli. This includes high-privilege blockchain operations such as broadcasting transactions, signing messages, and managing application security settings on the Alchemy platform.
[EXTERNAL_DOWNLOADS]: The skill instructs the agent to install and execute code from an external source using npm i -g @alchemy/cli or npx -y @alchemy/cli. These resources are downloaded from the vendor's official package registry.
[CREDENTIALS_UNSAFE]: The skill documents a procedure for extracting sensitive API keys using the --reveal flag and persisting them into a local .env file. While this is performed for standard project environment configuration, it involves the agent directly accessing and handling raw secrets.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process data from external blockchains and APIs. Maliciously crafted data stored on-chain could potentially influence the agent's processing logic. The skill includes boundary markers by mandating structured JSON output for all CLI interactions.
Ingestion points: CLI output from commands like evm data, solana das, and app list (SKILL.md).
Boundary markers: Explicitly requires alchemy --json --no-interactive to ensure structured data parsing (SKILL.md).
Capability inventory: Shell command execution, file system writing, transaction signing, and network operations (SKILL.md).
Sanitization: Relies on JSON parsing to handle external content; no explicit instruction for shell escaping of user-provided arguments.
[DYNAMIC_EXECUTION]: The skill contains instructions for a bridging flow that dynamically generates and executes a shell script to update environment variables using sed and local variables.

alchemy-cli