The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill instructs users to install the poke package from the npm registry. This is the official SDK required to interact with the service described in the documentation.
[DATA_EXFILTRATION]: The documentation describes the SDK's credential resolution logic, which includes reading authentication tokens from a local configuration file at ~/.config/poke/credentials.json. While this involves accessing sensitive data on the filesystem, it is a standard mechanism for CLI tools and SDKs to manage service-specific authentication.
[PROMPT_INJECTION]: The skill provides patterns for sending data (such as clipboard content, URLs, and code diffs) to an AI agent via sendMessage and webhooks. This creates a surface for indirect prompt injection if the processed data contains malicious instructions.
Ingestion points: The references/use-cases.md file explicitly mentions interpolating variables like selection, pageUrl, and diff into prompts sent to the agent.
Boundary markers: There are no instructions for using delimiters or 'ignore' instructions to isolate untrusted data within the prompt.
Capability inventory: The skill provides access to the sendMessage and sendWebhook methods for transmitting data to the Poke service.
Sanitization: No specific sanitization or validation logic is suggested for the external content before it is processed by the agent.

poke