How do I install the Model Census skill?

Run npx skillmds@latest add seansmithworks/model-census in your terminal (requires Node.js), paste this page's agent-chat prompt into Claude, Cursor, or any MCP-connected agent, or download the SKILL.md file and copy it into your agent's skills directory.

Is Model Census safe to use?

This skill has not completed SkillMD's automated safety review yet. SkillMD never runs a skill's scripts for you; review the SKILL.md before installing.

Which AI agents work with Model Census?

This skill is tagged as working with Claude Code, Claude.ai, OpenAI Codex. SKILL.md is an open format, so most agents that read a skills directory can load it too.

Is Model Census free to use?

Yes. Installing skills from SkillMD is free. This skill is licensed under MIT.

Who published Model Census?

seansmithworks (@seansmithworks) published this skill. Their other Agent Skills are listed on their SkillMD profile.

Model Census by seansmithworks — Agent Skill

Name: Model Census
Author: seansmithworks

Model Census

Report which Claude model actually ran, per main thread, per subagent type, and per repo, from local transcripts. Use to verify tier routing works and to see where Fable/Opus tokens go. Trigger: 'model census', 'model usage', 'which models are we using', '/model-census'.

seansmithworks 6604251 2 files · 9.1 KB Updated Sep 17, 2026

File contents

Model census

Neither the API console nor ccusage can see subagent type or repo — they see API keys and token counts, not which agent definition ran or what it was working on. This skill reads the transcripts directly.

What it reads

~/.claude/projects/**/*.jsonl (main-thread transcripts) and ~/.claude/projects/**/subagents/agent-*.jsonl (subagent transcripts). Subagent type is recovered by joining the parent transcript's Agent tool_use (input.subagent_type) to the matching tool_result's agentId: <id> line, then matching that id to its subagent transcript. Token counts are raw counts, not dollars — this is a routing check, not a billing report.

Command

python3 <skill dir>/scripts/model-census.py --brief --since <date>

Global install (npx skills add ... -g): ~/.claude/skills/model-census/scripts/model-census.py. Project-local install (the default): ./.claude/skills/model-census/scripts/model-census.py. --since takes YYYY-MM-DD (default: 30 days back). Drop --brief for the full report broken out by project.

Always flag

Any Fable or top-tier row in the MAIN THREAD table — the main thread pays to read every tool output, so a top-tier default there is the expensive place for it to leak, not a per-task subagent.
A top-level model key in ~/.claude/settings.json — that's a silent global default, usually written by /model <x> + Enter instead of the session-only flag.

Frequently asked questions