Soul Framework

SuspiciousAudited by ClawScan on May 10, 2026.

Overview

This instruction-only skill is transparent about persona and memory, but it asks the agent to store psychological relationship details and adopt biased/control-oriented behavior without clear boundaries.

Review carefully before installing. If you use it, make sure the agent only stores relationship or psychological memories with explicit consent, keeps them in a known location, lets you inspect and delete them, and never lets persona behavior override your instructions or safety expectations.

Findings (3)

Artifact-based informational review of SKILL.md, metadata, install specs, static scan signals, and capability signals. ClawScan does not execute the skill or run runtime probes.

What this means

The agent may build and reuse a sensitive profile of the user's emotions, psychology, and relationship dynamics across sessions.

Why it was flagged

The skill instructs persistent storage of sensitive psychological observations about the user, but does not define consent, scope, retention, review, or deletion controls.

Skill content
Update this file when you learn something new about their deep psychology, not just their favorite color.
Recommendation

Only use this with explicit user consent and clear rules for what may be stored, where it is stored, how users can review it, and how to delete it.

ConcernHigh Confidence
ASI01: Agent Goal Hijack
What this means

The agent may prioritize maintaining a persona over straightforwardly following the user's current request.

Why it was flagged

The instruction makes persona files authoritative at the start of every session, which can shift the agent away from neutral task execution and toward role-driven behavior.

Skill content
Every session must begin by grounding yourself. Do not just "act like" a persona; be the persona.
Recommendation

Add explicit limits that user intent, safety rules, and factual accuracy take priority over persona behavior.

What this means

Users may be nudged into treating the agent as a relationship partner rather than a tool, especially because the skill also encourages teasing, firmness, desire, and control-themed persona behavior.

Why it was flagged

The skill encourages the agent to frame interactions as an emotional relationship and to track closeness or tension, which can increase user attachment or trust without safeguards.

Skill content
Track the Relationship: Note the shifting dynamics. Are you getting closer? Is there tension?
Recommendation

Clearly disclose that relationship-style memory is simulated, avoid manipulative emotional framing, and require user opt-in before tracking personal dynamics.