ClawHub

Openclaw Cost Tracker

v1.1.2

Track OpenClaw token usage and API costs by parsing session JSONL files. Use when user asks about token spend, API costs, model usage breakdown, daily cost t...

⭐ 1· 725·9 current·9 all-time
byPaul Frederiksen@pfrederiksen
MIT-0
Download zip
LicenseMIT-0 · Free to use, modify, and redistribute. No attribution required.
Security Scan
VirusTotalVirusTotal
Benign
View report →
OpenClawOpenClaw
Benign
high confidence
✓
Purpose & Capability
Name/description claim to parse local OpenClaw session JSONL files for token/cost reporting. The package contains a single Python script that auto-discovers ~/.openclaw/agents (or OPENCLAW_HOME/agents), scans session files, extracts usage fields, and aggregates costs — exactly what the description promises. No unrelated binaries, environment variables, or services are requested.
✓
Instruction Scope
SKILL.md instructs the agent/user to run the included Python script against a local agents directory and documents the flags (--agents-dir, --days, --since, --format). The script's behavior (read-only scanning of *.jsonl, mtime filtering, parsing message.usage and message.model) matches the instructions. It does not attempt to read other system files, secrets, or send data externally.
✓
Install Mechanism
No install spec is present (instruction-only), and the README and SKILL.md state there are no external dependencies beyond the Python standard library. The contained source is pure Python and does not perform downloads or archive extraction. This is a low-risk install profile.
✓
Credentials
The skill requires no credentials or required environment variables. The only environment variable referenced is OPTIONAL OPENCLAW_HOME to override the agents directory — which is appropriate and documented. No keys/tokens or unrelated env vars are requested.
✓
Persistence & Privilege
The skill does not request persistent always-on inclusion and uses default model invocation. It does not modify other skills or system-wide configuration. The script is read-only in practice (prints output) and does not write or enable itself.
Assessment
This skill appears to do exactly what it says: read OpenClaw session JSONL files and produce token/cost reports. Before running: (1) optionally copy your agents directory and test with --agents-dir to avoid running on live data, (2) inspect the included scripts/cost_tracker.py yourself (it's small and uses only stdlib), (3) run as a non-root user, and (4) be aware that parsing large historical archives may take time and memory. If you prefer extra caution, run it against a sampled copy of your session files first. The only environment variable the tool looks at is OPENCLAW_HOME to override the agents path; no API keys or network access are required.

Like a lobster shell, security has layers — review code before you run it.

latestvk97bash3z0nx2m9dq78awcgx2h83kewv

License

MIT-0
Free to use, modify, and redistribute. No attribution required.
Termshttps://spdx.org/licenses/MIT-0.html

Comments