The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests large volumes of untrusted data from external sources, presenting a high risk of indirect prompt injection. Attacker-controlled data in smart contract logs or IPFS-hosted NFT metadata could contain instructions aimed at manipulating the agent's behavior.
Ingestion points: Processes blockchain logs and transaction data via RPC and Stream clients; retrieves and parses NFT metadata from IPFS as described in references/ipfs-reference.md.
Boundary markers: The instructions lack requirements for the agent to use delimiters or explicit 'ignore' warnings when processing retrieved external data.
Capability inventory: The agent has broad capabilities to modify infrastructure via the Admin API (references/admin-api-reference.md), execute local CLI commands (references/cli-reference.md), and initiate financial token swaps (references/swap-api-reference.md).
Sanitization: There are no explicit instructions for the agent to sanitize, escape, or validate content retrieved from the blockchain or IPFS before processing.
[COMMAND_EXECUTION]: The skill heavily utilizes the qn CLI tool for managing infrastructure. This gives the agent the ability to programmatically create, delete, and modify blockchain endpoints and data streams.
[EXTERNAL_DOWNLOADS]: The documentation instructs the agent to download and install various tools and libraries, including the @quicknode/sdk, mppx, and the qn CLI. These are official vendor-distributed resources from Quiknode Labs.
[DATA_EXFILTRATION]: The skill facilitates the retrieval of sensitive account information, including billing invoices, usage metrics, and team structures via the Admin API. It also provides capabilities for uploading data to IPFS and sending event data to external webhooks, which could be misused to exfiltrate sensitive local configuration files if the agent is compromised.

quicknode-skill