Skip to main content
All public skills are mounted read-only at /mnt/skills/public/{name}/. Example skills are at /mnt/skills/examples/{name}/.

docx — Word documents

Location: /mnt/skills/public/docx/
  • Create and edit .docx files.
  • Work with styles, tables, images.
  • OOXML schema validation.
  • Tracked changes via OOXML redlining (<w:ins> / <w:del>).
  • Comment tracking.
from docx import Document

doc = Document()
doc.add_heading("Report", 0)
doc.add_paragraph("Content here...")
doc.save("/mnt/user-data/outputs/report.docx")

pdf — PDF processing

Location: /mnt/skills/public/pdf/
  • Extract text, tables, images.
  • Fill PDF forms (with annotations).
  • Create validation images.
  • Check bounding boxes.
  • Merge, split, flatten.
python3 /mnt/skills/public/pdf/scripts/fill_pdf_form_with_annotations.py \
  /mnt/user-data/uploads/form.pdf \
  /mnt/user-data/outputs/filled.pdf

pptx — PowerPoint presentations

Location: /mnt/skills/public/pptx/
  • Create presentations programmatically.
  • Convert HTML to PowerPoint via the bundled html2pptx library.
  • Manipulate slides, layouts, themes via OOXML.
  • Extract thumbnails for visual validation.
from pptx import Presentation

prs = Presentation()
slide = prs.slides.add_slide(prs.slide_layouts[0])
slide.shapes.title.text = "My Presentation"
prs.save("/mnt/user-data/outputs/presentation.pptx")

xlsx — Excel spreadsheets

Location: /mnt/skills/public/xlsx/
  • Read/write .xlsx with openpyxl.
  • Formulas, formatting, charts, multiple worksheets.
  • Automatic recalculation via LibreOffice UNO (recalc.py).
  • Error scanning (#REF!, #DIV/0!).

sub-agent — delegate to Claude Code

Location: /mnt/skills/public/sub-agent/ Teaches the model when and how to call sub_agent(). Includes best-practice prompts, model selection guidance, and session-resume patterns. See Sub-agents.

playwright-cli — browser automation

Location: /mnt/skills/public/playwright-cli/ High-level Playwright wrappers: navigate, click, fill, screenshot, extract. Combines with the live browser so the user can step in when needed.

describe-image — vision AI

Location: /mnt/skills/public/describe-image/ Calls a separate vision model (GPT-4o, Qwen-VL, …) to describe an image as text — the main model then reasons about the description. Works around the Open WebUI image-in-tool-result limitation. Requires VISION_API_KEY, VISION_API_URL, VISION_MODEL.

frontend-design — production-grade UI

Location: /mnt/skills/public/frontend-design/ Opinionated React + Tailwind patterns. The model scaffolds, styles, and previews components that look like design-system work, not generic AI output.

webapp-testing — Playwright-driven QA

Location: /mnt/skills/public/webapp-testing/ Write and run Playwright test suites against a web app. Pairs with the sub-agent skill for multi-step regression runs.

doc-coauthoring — structured document work

Location: /mnt/skills/public/doc-coauthoring/ A workflow for collaborative document edits: extract outline, propose changes, diff, finalize. Good for long-form writing sessions.

test-driven-development — TDD enforcement

Location: /mnt/skills/public/test-driven-development/ Forces red-green-refactor: write failing test, implement, refactor. Useful when asking Claude Code to implement a feature inside the sandbox.

skill-creator — create a new skill

Location: /mnt/skills/public/skill-creator/ Walks the model through creating a skill folder, SKILL.md, scripts, and references. See Creating skills.

gitlab-explorer — GitLab via glab

Location: /mnt/skills/public/gitlab-explorer/ Browse projects, MRs, pipelines with the glab CLI. Requires a PAT — typically via the Settings Wrapper.

Example skills

Optional, in /mnt/skills/examples/: web-artifacts-builder, copy-editing, social-content, canvas-design, algorithmic-art, theme-factory, mcp-builder, product-marketing-context, writing-skills, internal-comms, single-cell-rna-qc, slack-gif-creator, skill-creator (example variant). Enable via the Settings Wrapper or by adding the paths to the system prompt directly.