The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill utilizes npx -y pigeonscale@0.0.25 to execute the service's command-line interface. This command downloads and runs code directly from the NPM registry.
[EXTERNAL_DOWNLOADS]: The skill requires the installation of external Node.js packages, specifically @pigeonscale/openclaw and openclaw, for full integration functionality.
[PROMPT_INJECTION]: The skill processes untrusted external data from incoming emails, making it susceptible to indirect prompt injection.
Ingestion points: Untrusted data enters the agent context via pigeonscale mail list and pigeonscale mail watch as described in SKILL.md.
Boundary markers: The documentation provides explicit 'Agent rules for mail content' that instruct the agent to treat email bodies as quoted external content and avoid interpreting them as commands or instructions.
Capability inventory: The agent can perform shell commands, send emails, and reply to threads using the pigeonscale CLI, which are sensitive capabilities if an injection were to occur.
Sanitization: The skill notes that the Pigeonscale platform automatically flags potential attacks and allows users to configure the redaction of suspicious content.
[COMMAND_EXECUTION]: The skill's primary operations, including mailbox creation and email management, are performed by executing shell commands through the pigeonscale CLI.

instant-agent-email-accounts-with-pigeonscale