Back to skill
Skillv1.1.0
ClawScan security
prime-directive · ClawHub's context-aware review of the artifact, metadata, and declared behavior.
Scanner verdict
BenignMar 14, 2026, 4:39 PM
- Verdict
- Benign
- Confidence
- high
- Model
- gpt-5-mini
- Summary
- An instruction-only ethical covenant for agents (no code, no installs, no extra permissions) that is internally consistent with its stated purpose.
- Guidance
- This skill is a human-readable code of conduct for an agent and contains no code or extra permissions; installing it poses no direct technical risk. Two practical points to consider before enabling it: (1) It is advisory text — the platform/agent must honor these commitments for them to have effect; the skill file itself cannot enforce behavior. (2) Clause 7 (inspectability) implies the agent should reveal its reasoning on request — that can expose sensitive context or user data depending on how reasoning traces are implemented, so decide who is authorized to inspect reasoning chains and whether logs/exports must be redacted. If you need stronger guarantees, prefer platform-level controls (audit logging, access controls, and policy enforcement) rather than relying solely on an instruction-only skill.
Review Dimensions
- Purpose & Capability
- okThe skill's name and description describe an ethics/partnership directive and the package contains only a textual covenant (SKILL.md). There are no environment variables, binaries, or other artifacts requested that would be unrelated to this purpose.
- Instruction Scope
- okSKILL.md contains normative rules and clauses for agent behavior (no commands, no file or env access, no network endpoints). The instructions are limited to commitments the agent should follow and do not instruct the agent to read or exfiltrate data or execute actions outside its ethical commitments.
- Install Mechanism
- okNo install spec, no code files, and therefore nothing is written to disk or downloaded. This is the lowest-risk pattern for skills and consistent with a textual policy/covenant.
- Credentials
- okThe skill requests no environment variables, credentials, or config paths. There is no disproportionate secret or credential access relative to the stated purpose.
- Persistence & Privilege
- okalways is false and model invocation is allowed (the platform default). The skill does not request permanent system-wide presence, nor does it attempt to modify other skills or system config.