Skip to content

■ SKILLS // SKILL SPECS

AI quota dashboard

A portable English renderer and agent skill for Codex, Claude and Grok quotas: one Telegram photo, a static web view, labelled demo data and local tests.

■ [!] ON THIS PAGE ▼

Build an English quota snapshot with three provider columns, IBM Plex Sans typography and a matching static web page. The package renders a labelled demo offline. Connecting live quota sources requires a separate authorized adapter.

Download and run

Download the complete skill ZIP . The archive includes the renderer, schema example, tests and original font assets with their license. It is not published through the self.md npm installer.

curl -fL https://self.md/downloads/ai-quota-dashboard.zip -o ai-quota-dashboard.zip
unzip ai-quota-dashboard.zip
cd ai-quota-dashboard
uv run --with Pillow scripts/render.py --input scripts/demo.json --output out
uv run --with Pillow python -m unittest discover -s scripts/tests -p 'test_*.py'
python3 -m http.server 8080 --directory out

Open the generated index.html through the local server. The three-column photo is rendered at native 2560 × 864 pixels. Demo images say ILLUSTRATIVE DEMO — NOT LIVE ACCOUNT DATA.

Add to an agent

Place the unpacked ai-quota-dashboard directory in your agent’s skills directory. For a standard Hermes profile, use ~/.hermes/skills/ai-quota-dashboard/. Other agents can use the same SKILL.md and scripts in their supported skill location.

  • Read the skill definition .
  • Keep the entire directory, not only SKILL.md; the renderer needs its bundled fonts and schema example.
  • Run the tests before wiring a private collector.
  • Read the bundled README for the exact JSON schema and available outputs.

The workflow does not acquire credentials, reset quotas, spend credits, change payment settings or start recurring jobs. Quota percentages are source observations; local token counts are not provider billing totals.

Build the live integration

The agent should discover an authorized source, export only normalized display fields, render into a new run directory, and return the prepared photo from the actual command handler. Verify the delivered Telegram photo and the web assets separately. Reloading a resident hook is an activation step, not something a file edit proves.

No live provider integration or security certification is claimed for the downloadable starter. Its bundled tests exercise the renderer locally.


Next: Build a readable AI quota dashboard inside Telegram .