Learning

• Outcome: Pass
• Latest audit: Jul 26, 2026 · 12:48 PM UTC
• Version: 1.0.3

Overview

This is an adaptive learning skill that keeps local study preferences and progress notes; the persistence creates privacy considerations but is coherent with the tutoring purpose and disclosed in the artifacts.

Install this if you want a tutor that remembers learning preferences, past misses, and topic progress between sessions. Review or clear the local files under ~/Clawic/data/learning/ if you do not want that history retained, especially on shared machines.

SkillSpector

By NVIDIA

Vulnerability Patterns

• Data Exfiltration: External Transmission, Env Variable Harvesting, File System Enumeration
• Excessive Agency: Unrestricted Tool Access, Autonomous Decision Making, Scope Creep
• MCP Tool Poisoning: Hidden Instructions, Unicode Deception, Parameter Description Injection
• Prompt Injection: Instruction Override, Hidden Instructions, Exfiltration Commands
• Privilege Escalation: Excessive Permissions, Sudo/Root Execution, Credential Access

Findings (4)

Description-Behavior Mismatch

Medium

• Confidence: 93% confidence
• Finding: The setup explicitly instructs the skill to persist learner preferences, behavioral hypotheses, and per-topic session logs across sessions, which goes beyond transient teaching needed for a single interaction. This creates a real privacy and data-minimization risk because the user is not told that profiling and long-term retention are occurring, and the retained data could reveal learning difficulties, interests, or behavioral patterns over time.

Context-Inappropriate Capability

Medium

• Confidence: 91% confidence
• Finding: The skill directs the agent to maintain long-term hypotheses about user behavior and learning preferences based on observed signals, then confirms them after repeated interactions. Even if intended to personalize teaching, this is effectively user profiling and can accumulate sensitive inferences without transparency or necessity proportional to the stated teaching function.

Missing User Warnings

Medium

• Confidence: 96% confidence
• Finding: The instructions tell the skill to read and write persistent config and memory files containing learner data without any user-facing warning, consent flow, or explanation of what is stored. Silent persistence materially increases privacy risk because users may believe they are in an ephemeral tutoring session while the system is retaining data behind the scenes.

Autonomous Decision Making

Medium

• Category: Excessive Agency

Content

- User declares a preference ("just show me code", "keep it short") → update the matching key in `~/Clawic/data/learning/config.yaml` immediately, without ceremony.
- Observed format signal (a correct generation after a format, a re-ask after a format) → hypothesis in `memory.md`; confirmed at 2 consistent signals (SKILL.md Rule 8); a contradicting signal resets the count.
- Session results (level placed, concepts covered, misses, retirements) → the topic log in `memory.md` every session. This is operational data, not preference — it needs no confirmation threshold.
- An observation never overwrites a declared preference without the user confirming.

If the user has said and shown nothing, store nothing beyond the topic log.

• Confidence: 83% confidence
• Finding: The skill is authorized to update persistent configuration immediately and to log session results automatically 'without ceremony' and in some cases 'with no confirmation threshold.' In context, the autonomy is not code-execution dangerous, but it is still a valid security/privacy concern because the agent is making lasting data-handling decisions on behalf of the user without explicit approval.

VirusTotal

64/64 vendors flagged this skill as clean.

Static analysis

No suspicious patterns detected.