The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill recommends using npx -y @grimoirelabs/cli, which downloads and executes unpinned code from a remote package repository at runtime. This allows for potential execution of arbitrary code if the package is updated or compromised.
[EXTERNAL_DOWNLOADS]: The skill fetches external packages @grimoirelabs/cli and @grimoirelabs/venues from the NPM registry during execution without version pinning or integrity checks.
[COMMAND_EXECUTION]: The skill relies on shell command execution via grimoire, npx, and bun to interact with market APIs and perform on-chain actions.
[CREDENTIALS_UNSAFE]: The skill manages highly sensitive credentials, including the HYPERLIQUID_PRIVATE_KEY environment variable and local keystore files for the withdraw command. Exposure of these credentials would grant full access to the user's financial assets.
[DATA_EXFILTRATION]: The withdraw command enables the transfer of assets to arbitrary destination addresses, representing a high-risk capability if parameters are manipulated by an attacker.
[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted market data (prices, order books, metadata) from the Hyperliquid API which is then processed by the agent. 1. Ingestion points: market data commands in SKILL.md. 2. Boundary markers: Absent. 3. Capability inventory: Fund withdrawal and shell command execution. 4. Sanitization: Absent. This creates a surface where malicious market data could potentially influence agent decisions.

grimoire-hyperliquid