agentsclimarketplace

Agentype

Skill cyzlmh/agentype

Local AI-agent usage analytics and persona archetypes

Install
npx -y skills add cyzlmh/agentype

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

One thing to look at

  • 2 stars2 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

What its author says it does

Copied from the file, not written here

Run the Agentype workflow for local AI-agent usage analysis: collect and cache deterministic JSON, infer a persona/archetype from aggregate usage signals, then render a terminal summary or PNG poster. Supports Claude Code, Codex, OpenCode, pi-agent, Gemini CLI, OpenClaw, Nanobot, and configured Nanobot-compatible roots. Use when the user asks to understand their agent usage, AI workflow, token footprint, preferred agents/models/projects, or "agentype".

SKILL.md

3.9 KB, as published. Nobody here has run it

Agentype

Agentype summarizes a user's local AI-agent history into a persona/archetype and usage overview.

When this skill is triggered, you MUST complete all four steps below. Do not stop after collecting stats, do not skip persona inference, and do not skip delivering the final poster or summary to the user.

When to Use

Use this skill when the user asks:

  • "what is my agentype?"
  • "analyze my agent usage"
  • "show my AI usage stats"
  • "which agents or models do I use most?"
  • "what persona am I based on my AI workflow?"
  • /agentype

Do not use it for billing estimates. Agentype reports tokens and local usage signals, not provider invoices.

What It Reads

Agentype collects local session and token metadata from supported agents where available:

  • Claude Code
  • Codex
  • OpenCode
  • pi-agent
  • Gemini CLI
  • OpenClaw
  • Nanobot
  • Nanobot-compatible JSONL roots configured through AGENTYPE_NANOBOT_ROOTS

Required Workflow (all four steps are mandatory)

The PyPI distribution is agentype-cli because agentype is not available on PyPI. The installed command is still agentype.

Step 1 — Collect stats

Run the CLI with --json-out to collect deterministic usage data and write it to output/agentype.json:

agentype --json-out

If agentype is not installed and there is no source checkout:

uvx --from agentype-cli agentype --json-out

From a source checkout:

uv run agentype --json-out

The CLI output at this point is raw stats only — it is not the final result. Continue to the next step.

Step 2 — Infer and fill the persona (agent-side, no CLI call)

Read output/agentype.json. From the aggregate signals — top projects, agents, models, skill metadata, token shape, and usage rhythm — infer the user's persona yourself. Then write these four top-level fields back into output/agentype.json, preserving all other fields:

  • archetype: short persona label (e.g. "Polyglot Automator").
  • description: one-line explanation of the archetype.
  • keywords: 3–6 concise keywords.
  • comment: 2–3 evidence-grounded sentences starting with "You are a...".

Step 3 — Render the filled JSON

Pass the updated file back to the CLI to produce the final formatted output:

agentype --json-in output/agentype.json

For chat, IM, or gateway environments that can display images, also generate the poster:

agentype --json-in output/agentype.json --png-out

Step 4 — Deliver to the user

  • Terminal agents: relay the full rendered text output (persona/archetype + top stats) directly to the user.
  • Chat or IM gateway agents: send a compact text summary and attach output/agentype.png.

Do not expose raw session files, prompts, private transcripts, or full JSON unless the user explicitly asks for debugging data.

Custom Local Paths

If the user's agent history lives outside default locations, configure AGENTYPE_NANOBOT_ROOTS before Step 1:

AGENTYPE_NANOBOT_ROOTS="/path/to/workspace:/path/to/another/root" agentype --json-out

For unsupported agent layouts, the collector paths live in src/agentype/paths.py and source adapters in src/agentype/sources/.

Debugging

If the user asks for debugging or validation, re-run Step 1 with -v and share the verbose output:

agentype -v --json-out

Keep looking

Skills are one crate of 328,083. Ordering is by how many stacks a row turns up in, so the top of any crate is what has actually been picked rather than what has the most stars.