The Agent Skills Directory

[COMMAND_EXECUTION]: The skill includes a standalone worker script (scripts/worker.mjs) designed to execute remote tasks. It uses node:child_process.spawn to run commands provided by the platform or configured by the user, facilitating the remote worker functionality.
[EXTERNAL_DOWNLOADS]: The skill communicates with the vendor's API at flow.nuanu.com to manage workspaces, retrieve task definitions, and perform authentication. These network operations are necessary for the skill's primary function as a platform integration.
[CREDENTIALS_UNSAFE]: The skill handles sensitive agent keys and enrollment tokens. It employs security best practices by storing these credentials in the user's config directory with restricted (0600) permissions and implements logic to redact secrets from log output.
[PROMPT_INJECTION]: The skill ingests instructions and context from the Nuanu Flow platform, creating an attack surface for indirect prompt injection.
Ingestion points: Task envelopes fetched from the vendor API in scripts/worker.mjs.
Boundary markers: Prompt construction in buildPrompt uses newline separation between system instructions and task context.
Capability inventory: The worker allows shell command execution via the spawn utility.
Sanitization: No explicit filtering or sanitization of remote instructions is implemented before processing.

nuanu-flow