Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx skills add pbc-os/smb-starter-kit --skill patent-figuregit clone --depth 1 https://github.com/pbc-os/smb-starter-kitWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/pbc-os/smb-starter-kit/patent-figure)<a href="https://agentmods.dev/skills/pbc-os/smb-starter-kit/patent-figure"><img src="https://agentmods.dev/badge/skills/pbc-os/smb-starter-kit/patent-figure/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/pbc-os/smb-starter-kit/patent-figure"><img src="https://agentmods.dev/badge/skills/pbc-os/smb-starter-kit/patent-figure.svg" alt="Reviewed on agentmods" width="80" height="20"></a>What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00043 | $0.01667 |
| Opus 5 | $0.00022 | $0.00834 |
| Sonnet 5 | $0.00009 | $0.00333 |
| Haiku 4.5 | $0.00004 | $0.00167 |
Grade A, and why
patent-figure scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 11d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 144 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Patent Figure Generator
Generates formal USPTO-style patent figure drawings from provisional patent application .md files, then iteratively refines them using targeted single-fix edits.
Prerequisites
nano-bananaskill installed (uses itsgenerate_image.pyandedit_image.pyscripts)GEMINI_API_KEYaccessible as an env var (seesecrets-managerskill for how to wire this from your secret store)pip3 install google-genai python-dotenvif not already installed
Core Workflow
v1 — fresh generation (no seed image):
Use generate_image.py with the full structured prompt (5 sections, see below).
v2+ — targeted iteration (previous image as seed):
Use edit_image.py with the previous version as seed + the same structured prompt + a "ONE FIX NEEDED" section.
Never use the Mermaid diagram as a seed. Mermaid diagrams use different internal labels than what the patent figure needs. The model will follow the Mermaid labels and override your prompt.
Step 1 — Parse the Provisional
From the provisional .md file, extract:
-
Brief Description — find
## BRIEF DESCRIPTION OF THE DRAWINGSand copy the line for the specific figure verbatim (e.g.,**FIG. 1** — System architecture overview: ...) -
Verify figure number — before generating anything, cross-check the figure number against the Brief Description. Users sometimes describe a figure by its content rather than its actual number (e.g., asking for "FIG. 2 — the trust tier diagram" when FIG. 2 is actually the signal pipeline and FIG. 3 is the tier diagram). Always generate the figure that matches the spec for that number, and note any discrepancy to the user.
-
Reference numerals — scan the
## DETAILED DESCRIPTIONsection for every component with a parenthetical number (e.g.,Discovery Engine (210)). Build an exact table:210 = DISCOVERY ENGINE. These must match what will appear in the figure exactly — this table is the single most important section of your prompt and prevents the model from hallucinating component labels. If the provisional has no parenthetical numerals yet, assign them yourself starting from the 100-series for top-level components, 200-series for platform internals, 300-series for external components.
What ships with it
3 files beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 11d ago First seen · 144 lines · 43 tokens per session scan A 307e3d80a82d
patent-figure is a skill published in the GitHub repository pbc-os/smb-starter-kit (10 stars, last pushed 3mo ago), licensed MIT. It adds 43 tokens to every session and 1,667 once invoked, about $0.0002 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other skills, from other repositories
seedance-copyright
This skill should be used when a Seedance 2.0 prompt mentions named characters, franchises, studios, celebrities, public figures, private people, brand logos, copyrighted scenes, songs, voices, or real-person likeness workflows and needs an IP-safe rewrite.
voice-conversion-studio
A voice-conversion tool that changes a provided recording into an authorised target voice.
voice-clone-lab
A consent-controlled tool for creating a reusable synthetic voice from a local speaker recording.
c2pa-metadata
Embed a C2PA provenance manifest into an AI-generated marketing asset (PNG, JPG, WebP, GIF, TIFF, MP4, MOV, WebM, MP3, WAV, PDF) via scripts/embed-c2pa.py — produces a signed copy of the file carrying IPTC digital-source-type AI claims, an optional c2pa.ai-disclosure assertion for EU AI Act Article 50 (applicable 2…
suede-rights-passport
Suede Labs skill that turns messy creator materials into a local, offline rights-and-provenance transfer package: inventoried and hashed assets, a normalized suede-intake.json manifest, credits and splits, license notes, provenance, and a missing-information report, validated by a bundled stdlib script. Use when a…
suede-rights-audit
Suede Labs skill that finds and organizes the rights gaps in a creator project before packaging: ownership, contributors, splits, samples, licenses, provenance, metadata, licensing readiness, and royalty-routing readiness, each marked confirmed or unknown against an evidence trail. Use when a song, release, or…