The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses the cast CLI tool (part of the Foundry toolkit) to perform blockchain operations such as querying balances, checking allowances, and broadcasting transactions.
Evidence: The skill relies on multiple shell utilities including cast, jq, bc, sed, printf, and date within references/evm-cli.md to automate transaction construction and receipt verification.
Guardrails: It implements mandatory user confirmation prompts (Reply exactly: YES) and prioritizes browser-based signing (--browser), which ensures private keys are not exposed to the agent or terminal.
[EXTERNAL_DOWNLOADS]: The skill references external tools and installation scripts from well-known sources.
Evidence: SKILL.md suggests installing additional skills via npx from the sablier-labs repository. references/evm-cli.md directs users to https://getfoundry.sh/ to install the required CLI tools.
[PROMPT_INJECTION]: The skill processes user-provided data such as token addresses and vesting details which are subsequently used in shell command arguments.
Ingestion points: SKILL.md defines arguments for chain_name, vested_token_address, and vesting_details.
Boundary markers: The skill does not explicitly use delimiters (like triple quotes or XML tags) to wrap user input in its internal prompts.
Capability inventory: The skill possesses significant capabilities through shell execution (cast send, etc.) as documented in references/evm-cli.md.
Sanitization: The workflow requires the agent to manually validate inputs, infer intent, and map descriptions to specific ABI functions, which provides a layer of logic-based sanitization, though explicit shell escaping is not mentioned.

sablier-create-vesting