Prompt Token Counter
PassAudited by ClawScan on May 1, 2026.
Overview
This looks like a coherent token-counting and audit skill, with disclosed local OpenClaw file access and optional URL/API benchmark features that should be used deliberately.
Install if you are comfortable with a local token counter reading selected OpenClaw workspace and skill files for audit results. Avoid using URL fetching on untrusted links, and use the benchmark API mode only if you intentionally want to call a provider API with your own key.
Findings (4)
Artifact-based informational review of SKILL.md, metadata, install specs, static scan signals, and capability signals. ClawScan does not execute the skill or run runtime probes.
The local token counter may process private OpenClaw memory/persona/skill text to compute token counts.
The primary audit mode reads persistent OpenClaw memory, persona, and skill files. This is disclosed and purpose-aligned, but users should notice it because those files can contain sensitive private context.
The audit workflow reads files under `~/.openclaw/workspace` and `~/.openclaw/skills` ... Those files may contain personal data or secrets.
Run audits only on files you intend to inspect, and avoid storing secrets in OpenClaw memory or skill files.
If used with an untrusted URL, the machine running the skill will contact that URL.
The CLI can make outbound HTTP(S) requests when given a URL. The skill documentation requires user confirmation, so this appears disclosed and optional, but untrusted URLs can still expose IP/network metadata or cause SSRF-like risks.
if parsed.scheme not in {"http", "https"}: ... urllib.request.urlopen(url, timeout=10)Prefer local files or inline text; use --url/-u only for URLs you explicitly trust and have confirmed.
If the benchmark API mode is run, provider credentials may be used and requests may be sent to the configured model API.
The optional benchmark script may use provider API credentials and make model API calls. This is documented as a benchmark mode rather than required for the main local counter.
**API mode** (default): Uses model API to get exact `prompt_tokens`. Set `API_KEY` and `BASE_URL` in the script.
Use local mode for benchmarking unless you intend to call a provider API; use limited, revocable keys and avoid embedding sensitive long-lived keys in example scripts.
Users have less registry-level provenance information to confirm they are installing the intended package.
The registry metadata does not populate source or homepage fields, which is a provenance gap even though package documentation separately names a GitHub repository.
Source: unknown; Homepage: none
Verify the package source and repository before installation, especially if installing from npm or a non-registry copy.