Skillv0.1.3

ClawScan security

Token Cost Time · ClawHub's context-aware review of the artifact, metadata, and declared behavior.

Scanner verdict

BenignApr 18, 2026, 12:04 AM

Verdict: benign
Confidence: high
Model: gpt-5-mini
Summary: The skill's code, instructions, and file I/O are consistent with its stated purpose (local token/cost/duration estimation and local profiling); it does not request external credentials or network endpoints, but it does persist profiling data to your home directory so review where and what it writes before use.
Guidance: This skill is internally consistent and runs entirely locally, but before installing: 1) Review and accept that it will write two files under your home directory (~/.token-cost-time/profile.json and execution-log.jsonl). Those logs contain model names, token counts, durations, costs, retries, and timestamps (they do not include prompt text or model outputs). 2) If you wire automation (cron or agent hooks) to call record(), ensure that automation does not capture or write sensitive prompt or output content into these logs. 3) The package.json in the bundle lists version 0.1.0 while registry metadata is 0.1.3 — consider verifying source/origin and integrity if provenance matters. 4) Because this runs as Node code, inspect the shipped files yourself or run in an isolated environment if you have heightened security/privacy requirements.

Review Dimensions

Purpose & Capability: okName/description match the actual code: rule-based classification, baseline priors, cost calculation, local recording, and a CLI. The files and functions provided (calibrate, record, compare) are exactly what the README claims. No unrelated binaries or credentials are requested.
Instruction Scope: noteSKILL.md and the CLI limit operations to classification, estimation, and local recording. The README suggests wiring a cron that reads 'session_status' to auto-record runs — that integration may read other session metadata in your environment. The skill itself does not collect prompts or outputs, only model name, token counts, duration, cost, retries, and timestamps. Confirm any automated integration does not forward sensitive prompt/content to the logs.
Install Mechanism: okNo install spec or network downloads; this is a local JS module intended to run with Node (>=18). All source files are included and use Node built-ins only. No remote code fetching or archive extraction is present.
Credentials: okThe skill requires no environment variables, no credentials, and no special config paths beyond writing to files under the user's home directory (default: ~/.token-cost-time/). That file access is proportional to the purpose (local profile + execution log).
Persistence & Privilege: okDoes not request always:true or elevated privileges. It persists its own profile and log under ~/.token-cost-time and does not modify other skills or system-wide agent configuration.