The Agent Skills Directory

[UNVERIFIABLE_DEPENDENCIES_AND_REMOTE_CODE_EXECUTION]: The skill provides instructions to install the vendor's own CLI tool from the NPM registry using 'npm install -g dineway@latest'. This is standard setup for the intended functionality.
[DATA_EXPOSURE_AND_EXFILTRATION]: A workflow for parsing local files is described (Flow 3) which involves encoding file content into Base64 and sending it to a remote provider endpoint ('context.dev'). This is a functional feature of the service and includes explicit warnings to avoid passing credentials through the command or conversation.
[INDIRECT_PROMPT_INJECTION]: The skill processes data from various third-party API providers, presenting a potential surface for indirect prompt injection.
Ingestion points: Data is retrieved from external providers through the 'dineway tools run' and 'dineway tools runs get' commands in 'SKILL.md'.
Boundary markers: The skill mandates a 'discover then inspect' workflow, requiring the agent to review provider-owned schemas and risk policies before execution.
Capability inventory: The agent can execute the 'dineway' CLI and save outputs to the filesystem using the '--output' flag.
Sanitization: Instructions specify mapping inputs strictly to the flags ('--query', '--path', '-i') defined by the inspected JSON Schema contracts to ensure data is handled as structured parameters.

dineway-tools