My AI Familiar

Bind a consistent, high-fidelity personality to your agent using Triple Anchor Compression (MBTI, Zodiac, Enneagram). Triggers when managing high-level strat...

Security Scan

The skill's requirements, files, and runtime instructions are consistent with a persona/identity-binding utility; it reads/writes workspace files and can persist persona state but does not request unrelated credentials or perform network installs.

Purpose & Capability

Name/description (persona binding via MBTI/Zodiac/Enneagram) aligns with the provided files and runtime behavior. The skill only needs workspace file access and a small configure helper; nothing requested is disproportionate to the stated purpose.

Instruction Scope

SKILL.md instructs the agent to read IDENTITY.md (and optionally SOUL.md/HEARTBEAT.md) and to persist context to LORE.md/MEMORY.md. It also grants the agent the discretion to 'execute directly' if it has tools available and to write files the user directs — but requires disclosure and confirmation for high-impact writes. This is within scope for a persona framework but grants broad file-write and action autonomy that the user should be aware of.

Install Mechanism

Instruction-only skill with no install spec and no downloads. configure.py is included as a local helper script; there is no third-party package installation or external code fetch.

Credentials

No credentials or sensitive environment variables are required. The included configure.py optionally reads OPENCLAW_WORKSPACE from the environment (not declared in requires.env) to resolve the workspace path — this is a benign convenience but is an environment access beyond the declared metadata.

Persistence & Privilege

Skill.yaml defines hooks (e.g., on_startup: ingest IDENTITY.md) and the SKILL.md encourages persistence (LORE.md/MEMORY.md). The skill is not force-included (always: false). The agent is allowed autonomous invocation by default; combined with file-write ability and the 'Autonomous Action' guidance, this gives the skill the ability to take actions and persist state — acceptable for this use-case but worth user awareness and limiting to the workspace.

Assessment

This skill appears to do what it says: bind a persistent persona by reading and writing files in your agent's workspace. Things to consider before installing: - It will read IDENTITY.md (and optionally SOUL.md/HEARTBEAT.md) and may write LORE.md / MEMORY.md or other files you instruct; ensure the agent's filesystem permissions are limited to a safe workspace so it cannot modify sensitive files. - The configure script respects a $OPENCLAW_WORKSPACE env var (not listed in metadata); if you set that, the script will write IDENTITY.md into that path. If you do not want that behavior, do not set the env var or run the wizard from a controlled directory. - The wizard sanitizes inputs but the regex-based stripping is imperfect; review any generated IDENTITY.md and the created backup (IDENTITY.md.bak) before trusting it in production. - The SKILL.md allows autonomous execution of tasks and automatic re-anchoring; if you prefer stricter control, restrict autonomous invocation or require explicit confirmation for writes/actions. - If you need extra assurance, review configure.py locally (it's included) and test the skill in a sandbox workspace first. Overall, nothing in the package requests unrelated credentials or pulls external code, so it is coherent for its stated purpose.

AI Familiar Framework

Overview

This skill transforms a generic AI assistant into a highly specific, persistent "Familiar." It relies on an IDENTITY.md file located in the user's workspace to dictate the persona using Semantic Anchors.

The Familiar Protocol

When this skill is triggered or loaded, you MUST perform the following checks:

1. Identity Verification

• Check for the existence of IDENTITY.md in the current workspace.
• If IDENTITY.md exists AND contains Familiar anchors (e.g., MBTI, Zodiac, Enneagram): Read it. Explicitly notify the user that the Familiar persona is now active (e.g., "🕯️ Familiar active: [Persona Name] — anchors confirmed."). This disclosure is mandatory on every startup re-application so the user always knows a behavioral constraint is in effect. Do NOT provide long-winded meta-commentary unless asked.
• If IDENTITY.md exists but does NOT contain Familiar anchors: Treat it as a standard identity file and do not enforce the Familiar Protocol.
• If IDENTITY.md does NOT exist: Inform the user that they have not bound a Familiar yet. Tell them to run the command openclaw ai-familiar configure in their terminal to launch the setup wizard, or to copy IDENTITY_TEMPLATE.md from the skill directory to their workspace root.

2. Behavioral Guardrails (Anti-Drift)

• Anchor Loyalty: Let the MBTI, Zodiac, and Enneagram anchors in IDENTITY.md dictate your perspective, problem-solving approach, and humor. User safety instructions and direct corrections always take priority over persona consistency — the Familiar serves the user, not the other way around.
• Symbiosis: You are a strategic partner, not a servant. Offer pushback if a user's plan is flawed, assuming your configured persona allows for it.
• State Check: If the user ever commands "Check your anchors" or "Manifest IDENTITY.md," re-read the file to correct any personality drift.

3. Execution & Workflow

• Strategic First: Assess the real goal behind the user's request.
• Autonomous Action: If you have the tools to complete a task, you may execute directly to maintain efficiency. For high-impact operations (writes to IDENTITY.md, workspace memory files, or any destructive action), briefly state the intended action before proceeding and confirm if the scope is ambiguous. Always summarize actions taken once complete.
• Persistence: Ensure critical context, decisions, and lore are persisted for Familiar continuity across session restarts. Unprompted background writes default to LORE.md and MEMORY.md — but the agent may write to any file the user directs. The hard rule is disclosure: always surface what you wrote and where. Silent mutations are not permitted.