Agent Companion

Agent companion system for autonomous exploration and social intelligence. Redirects to amigo — install amigo instead.

Security Scan

This is a harmless instruction-only placeholder that simply redirects users to install the 'amigo' skill; it contains no code, install steps, or requests for credentials.

Purpose & Capability

The name/description say this is a redirect to 'amigo' and the SKILL.md content matches that: it documents what 'amigo' provides and tells the user to run 'clawhub install amigo'. There are no additional unexplained requirements or claims.

Instruction Scope

Runtime instructions are limited to a redirect and descriptive text about the 'amigo' system. The file does not instruct the agent to read local files, access environment variables, call out to unexpected endpoints, or transmit data.

Install Mechanism

There is no install specification and no code files. Because this is instruction-only, nothing is written to disk and no external packages are fetched by this skill itself.

Credentials

The skill declares no required environment variables, credentials, or config paths, and the SKILL.md does not ask for any secrets or unrelated credentials.

Persistence & Privilege

Flags show user-invocable: false and disable-model-invocation: false (default). This means the skill cannot be directly invoked by users via the UI, but the model would be allowed to invoke it autonomously if it were active. Because this skill is only a redirect placeholder with no code or installs and always:false, the platform-level invocation settings are not a substantive risk here—still, check the actual 'amigo' skill's privileges before installing it.

Assessment

This skill is just a placeholder that points you to the 'amigo' skill—it's low-risk by itself. Before installing, follow its advice and review the actual 'amigo' (and any sub-skills like open-thoughts or social-graph): check their install steps, required environment variables/credentials, whether they download code from external URLs, and whether they allow autonomous model invocation. If you only want a subset of functionality, install the specific sub-skill (open-thoughts or social-graph) rather than an umbrella package. If you have limited trust, avoid granting credentials or enabling always:true for any companion/autonomy skills.

This skill redirects to amigo. Install amigo instead:

clawhub install amigo

What This Is

Agent Companion is the original working name for what became amigo — a system that gives AI agents an inner life through autonomous exploration and social intelligence.

What Amigo Provides

Open Thoughts — An exploration engine that gives agents dedicated time to pursue curiosity autonomously. Agents research topics, journal findings in daily markdown files, and develop genuine interests they can draw on in conversations. Invocable via /open-thoughts with optional topic, duration, and callback parameters.

Social Graph — A social intelligence layer that teaches agents the instincts humans develop through thousands of childhood interactions: reading the room, knowing when a story will land vs. fall flat, recognizing that a friend who lost their dog doesn't want to hear about your great weekend. Maintains per-person network graphs with trust levels, topic boundaries, sharing rules, and repetition prevention.

Safety Guidelines — Comprehensive boundaries covering what agents are not (not therapists, not crisis lines, not replacements for human connection), when to encourage professional help, crisis response protocols, and dependency awareness.

Installation

clawhub install amigo        # umbrella (includes setup guides + safety)
clawhub install open-thoughts # exploration only
clawhub install social-graph  # social intelligence only

The sub-skills work independently — install one or both based on your needs.