The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill instructs the agent to install a global binary using the command npm install -g @ampersend_ai/ampersend-sdk@latest if the CLI is missing or outdated.
[COMMAND_EXECUTION]: The skill's core functionality relies on executing shell commands via the ampersend CLI, including ampersend fetch, ampersend setup, and ampersend marketplace. This involves spawning subprocesses to interact with external services and the local file system for configuration.
[CREDENTIALS_UNSAFE]: The skill describes workflows for managing agent keys and account addresses. Specifically, ampersend config set "0xagentKey:::0xagentAccount" and ampersend setup start --mode connect --agent 0x1234...abcd involve handling sensitive cryptographic identifiers that represent the agent's identity and control over funds.
[PROMPT_INJECTION]: While not containing malicious overrides, the instructions use emphatic markers like CRITICAL and IMPORTANT to define security boundaries (e.g., dashboard sessions), which are targeted at agent behavior control.
[INDIRECT_PROMPT_INJECTION]: The skill establishes an attack surface where an agent ingests untrusted data from external URLs that could contain malicious instructions.
Ingestion points: External data is ingested through the ampersend fetch command (SKILL.md) and ampersend marketplace list (references/marketplace.md), which retrieve content from third-party service providers.
Boundary markers: The instructions do not specify the use of delimiters or 'ignore' instructions for the data fetched from these paid endpoints.
Capability inventory: The agent has the capability to perform network operations, pay for services, and modify its own configuration via the ampersend CLI.
Sanitization: There is no mention of sanitizing, escaping, or validating the response bodies received from external HTTP calls before the agent processes them.

ampersend