doc-skills — Document-Type Skill Collection
Author: Simon-Pierre Boucher Contact: contact@spboucher.ai
Ten ultra-sharp skills, one per document type, each covering create, read,
and modify with a single default tool per operation, an explicit escape
hatch, a validation step, and a references/recipes.md for deep recipes.
All skills follow the principles in ../RESEARCH-SYNTHESIS.md
and pass python3 ../tools/validate_skills.py.
The collection and its boundaries
| Skill | Handles | Explicitly does NOT handle |
|---|---|---|
processing-xlsx |
.xlsx/.xlsm/.xltx spreadsheets | .csv/.tsv → processing-csv; profiling |
processing-csv |
.csv/.tsv create/read/edit/convert | quality audits → profiling-csv-data; Excel files |
processing-docx |
Word .docx/.dotx | PDFs, spreadsheets, Google Docs, Markdown |
processing-pptx |
PowerPoint .pptx/.potx | Word docs, PDFs, Google Slides |
processing-pdf |
.pdf read/create/merge/split/forms | Word docs, images, Office→PDF export |
processing-json |
.json/.jsonl create/read/edit/validate/query | YAML/TOML configs; API design |
processing-yaml |
.yaml/.yml incl. configs | JSON files; CI pipeline logic |
processing-xml |
.xml create/read/edit/validate/XPath | HTML pages; OOXML internals of Office files |
processing-markdown |
.md create/read/edit/convert | release-notes house style; HTML files |
processing-html |
.html create/read/extract/edit | live web fetching; XML data files |
Boundaries are deliberately pairwise-exclusive so that no user request can plausibly trigger two skills at once — the #1 defense against false-positive triggering in a large collection.
Shared conventions
- Frontmatter description = WHAT + "Use when …" (literal phrases, extensions) + "Do not use for …"
- One default library per operation; escape hatch named with its install command
- Every write is followed by a re-parse/re-open validation step
- Malformed input → report the parser's error verbatim; never guess or regex-patch
SKILL.md<150 lines; depth lives inreferences/recipes.md(with TOC)