Back to skill
Skillv1.0.0
ClawScan security
Data quality & reconciliation with exception · ClawHub's context-aware review of the artifact, metadata, and declared behavior.
Scanner verdict
BenignMay 1, 2026, 4:50 AM
- Verdict
- Benign
- Confidence
- high
- Model
- gpt-5.5
- Summary
- This instruction-only skill appears benign: it provides a read-only workflow for reconciling user-provided datasets and producing exception reports, with no code, install step, credentials, persistence, or hidden data flow.
- Guidance
- The skill is safe to use as a structured reconciliation guide, but the datasets may contain personal or employment identifiers. Provide only the files needed for the task, confirm matching rules and tolerances, and review exception reports before making changes in any source system.
Review Dimensions
- Purpose & Capability
- okThe stated purpose and instructions are coherent: reconcile two or more CSV/XLSX datasets using stable identifiers, categorize mismatches, and produce exception reporting.
- Instruction Scope
- okThe workflow is bounded and includes user-control safeguards such as asking the user when columns, ID priority, or tolerances are unclear.
- Install Mechanism
- okThere is no install specification and no executable code; the included files are a rules reference and a CSV report template.
- Credentials
- okThe skill expects user-provided datasets containing pay numbers or driver document numbers, which is sensitive but proportionate to the reconciliation purpose and not paired with network access, credentials, or hidden sharing.
- Persistence & Privilege
- okThe artifacts do not request credentials, config paths, background execution, persistent memory, privileged access, or mutation authority.