Back to skill
Skillv1.1.1
ClawScan security
[Nyx Archive] Game Design Philosophy · ClawHub's context-aware review of the artifact, metadata, and declared behavior.
Scanner verdict
BenignFeb 23, 2026, 7:15 PM
- Verdict
- Benign
- Confidence
- high
- Model
- gpt-5-mini
- Summary
- The skill's claimed purpose (learning and adapting to a designer's instincts) matches its instructions and requested footprint; the main concern is privacy/vagueness around its 'always-on' passive learning behavior rather than any incoherence or unnecessary privileges.
- Guidance
- This skill is coherent with its description: it passively learns your game-design preferences from conversation and uses that to tailor advice. Before installing, consider privacy implications: 'observe and note' and 'always on' are vague — ask how long preferences are retained, whether data from unrelated chats could be used, and how to opt out or reset learned preferences. Because it's instruction-only and requests no secrets or installs, there's low technical risk, but review your agent's data-retention/permissions settings if you don't want personalization to persist across sessions.
Review Dimensions
- Purpose & Capability
- okThe name/description (adaptive game-design philosophy) aligns with the SKILL.md instructions: observe conversational signals about design preferences and apply them when giving guidance. There are no unrelated requests (no credentials, binaries, or config paths) that would contradict the stated purpose.
- Instruction Scope
- noteInstructions explicitly ask the agent to 'observe and note' design-related aspects of user conversation and then apply those learnings to future project work. That behavior is consistent with personalization, but the language is broad/vague (e.g., 'observe and note', 'always on'), which could result in harvesting context from unrelated conversations or retaining sensitive details unless constrained by agent policies.
- Install Mechanism
- okInstruction-only skill with no install spec and no code files; nothing is written to disk and no third-party packages are pulled in. Low install/remote-code risk.
- Credentials
- okSkill declares no required environment variables, credentials, or config paths. The requested data (conversation context) is appropriate for a personalization skill; no disproportionate secret access is requested.
- Persistence & Privilege
- notealways is false (not force-included) and autonomous invocation is allowed (the platform default). The SKILL.md's 'Passive Learning (Always On)' phrasing implies persistent behavioral adaptation across conversations, which is a privacy/retention concern but not an incoherence with its purpose. Users should verify platform data-retention and opt-out controls.