← Back to Skills Marketplace
Token Usage
by
PanthroCorp
· GitHub ↗
· v0.2.2
· MIT-0
181
Downloads
0
Stars
0
Active Installs
3
Versions
Install in OpenClaw
/install panthrocorp-token-usage
Description
Multi-agent token burn analysis across all registered OpenClaw agents
Usage Guidance
This skill is coherent with its purpose: it reads OpenClaw agent session files (~/.openclaw/agents/*/sessions/sessions.json), produces aggregated reports, and appends one history line to ~/.openclaw/workspace/memory/token-usage-history.ndjson. It does not request credentials, install software, or send data externally. Before installing or allowing autonomous use, decide whether you are comfortable with the skill creating/appending the local history file (it will do so on every run). If you want to be cautious: (1) inspect the sessions.json files to confirm they don't contain unexpected sensitive data, (2) run the skill manually once rather than allowing autonomous invocation, or (3) disable autonomous invocation for the skill so it cannot run without your explicit request.
Capability Analysis
Type: OpenClaw Skill
Name: panthrocorp-token-usage
Version: 0.2.2
The skill is a utility designed to analyze and report on token usage and costs across OpenClaw agents. It reads local session data from `~/.openclaw/agents/` and writes history/logs to the local workspace memory. The instructions in `SKILL.md` are focused entirely on data aggregation and reporting, with no evidence of network access, credential theft, or unauthorized execution.
Capability Assessment
Purpose & Capability
The skill's name/description align with its actions: it enumerates ~/.openclaw/agents/, reads sessions/sessions.json, aggregates token and cost fields, and reports summaries. There are no unrelated required binaries, env vars, or installs.
Instruction Scope
Instructions are narrowly scoped to files under the user's OpenClaw home (~/.openclaw). The only non-reporting action is appending one JSON line to ~/.openclaw/workspace/memory/token-usage-history.ndjson (persistence described in Step 7). The SKILL.md explicitly states that the append runs every time the skill runs — this is expected for a history feature but worth noting.
Install Mechanism
Instruction-only skill with no install steps and no code files — minimal surface area and nothing new is written to disk beyond the declared history file append.
Credentials
The skill requests no environment variables, no credentials, and no external config paths beyond the OpenClaw workspace; required access is proportional to the stated task.
Persistence & Privilege
The skill persists a single NDJSON record to ~/.openclaw/workspace/memory/token-usage-history.ndjson on every run. always:false (not force-enabled) and no external network activity are good, but because model invocation is allowed by default, an autonomously-invoked run could write this history without a separate explicit user prompt.
How to Use
- Make sure OpenClaw is installed (local or Docker)
- Run the install command in chat:
/install panthrocorp-token-usage - After installation, invoke the skill by name or use
/panthrocorp-token-usage - Provide required inputs per the skill's parameter spec and get structured output
Version History
v0.2.2
**Adds persistent structured usage history and data export**
- Appends every run's analysis as a compact JSON line to `token-usage-history.ndjson` for historical tracking (runs automatically).
- Specifies new line-by-line schema: per-agent, per-category stats and combined totals, with explicit nulls where data is absent.
- Clarifies that logging to `token-diet-log.md` is still user-optional.
- Updates permissions to reflect new always-on write for the history file.
- Bumps version to 0.2.2.
v0.2.1
panthrocorp-token-usage 0.2.1
- Updated skill instructions and documentation with improved example values and formatting in SKILL.md.
- Clarified sample session keys and agent names to be more generic and representative.
- Made minor text and formatting updates to documentation for improved consistency.
v0.2.0
- Initial release of Token Usage analysis skill for OpenClaw agents.
- Analyzes and aggregates token usage and costs across all registered agents.
- Supports dynamic agent discovery, grouping by session category, time window filtering, and anomaly flagging.
- Presents per-agent breakdown tables with category-level stats and flags for high-cost and suspected wasted usage.
- Provides a combined total summary and top 3 categories by cost and token volume.
- Optionally writes a concise Markdown log report when requested.
Metadata
Frequently Asked Questions
What is Token Usage?
Multi-agent token burn analysis across all registered OpenClaw agents. It is an AI Agent Skill for Claude Code / OpenClaw, with 181 downloads so far.
How do I install Token Usage?
Run "/install panthrocorp-token-usage" in the OpenClaw or Claude Code chat to install it in one step — no extra setup required.
Is Token Usage free?
Yes, Token Usage is completely free, licensed under MIT-0. You can download, install and use it at no cost.
Which platforms does Token Usage support?
Token Usage is cross-platform and runs anywhere OpenClaw / Claude Code is available (linux).
Who created Token Usage?
It is built and maintained by PanthroCorp (@panthrocorp); the current version is v0.2.2.
More Skills