Shapes a pile of raw notes into an article paragraph by paragraph, starting from competing openings and arguing about each block's form.
A thoughtful, opinionated editing loop, but it is marked in-progress upstream and demands a lot of back-and-forth.
writing-shape is an in-progress skill for the exploit stage of writing, when the raw material is fixed and you are committing to a structure. The agent reads your markdown file end to end (tidy fragments, loose prose or a transcript all work), treats it as read-only, and settles with you what the reader already knows. It then drafts two or three candidate openings, each implying a different thesis, and makes you pick one or combine them.
After that it grows the article block by block. For each step it asks what the reader needs to hear next, pulls material from the pile, and argues the form: prose or list, callout or inline, table or bold leads, quote or paraphrase, code block or inline code. Each agreed block is appended to the article file immediately, with the file re-read from disk first so your edits survive. A grounding rule means no block may lean on a concept the reader has not met. If the pile lacks an example, it names the gap and asks for one or cuts the section. The tone is a pushy editor ('If I cut this, what breaks?'). It is user-invoked, in a single 6.5 KB SKILL.md.
A single 6.6 KB SKILL.md describing a loop and conversational moves; no references, scripts or evals.
We scanned all 2 files in this skill folder and read every script. Nothing in it reaches beyond what the skill describes.
A user-invoked writing skill that reads a markdown file you supply (read-only) and builds an article file paragraph by paragraph at a path you choose. It has no scripts, network references or tool pre-approvals.
The skill folder (2 files) on Oct 29, 2026. We had every file and read every script.
NVIDIA SkillSpector static scan: 0 pattern hits. SkillSpector reported 0 hits. SKILL.md and agents/openai.yaml were read in full; they hold prose instructions and a display name plus an implicit-invocation-off policy.
A review, not a certification. Skills change after we check them — read the files before installing.
MIT (repo-level license on mattpocock/skills) — commercial use allowed.
Not really. The workflow depends on constant author decisions.
A lightweight CN/EN academic-writing skill that restructures paper drafts around one strongest claim instead of hedged, evenly-weighted self-critique.
A Korean-language humanizer that detects and rewrites AI-sounding text (translationese, mechanical parallelism) while preserving facts, with deterministic gates.
It guides an interactive session that turns raw notes into an article: you choose an opening, then the agent proposes and argues over each following paragraph or block, appending agreed text to a separate file.
Install the mattpocock/skills pack and call the skill with a path to your notes. It sits under skills/in-progress in the repo, so confirm your installer includes that folder.
writing-fragments is the explore step, collecting raw fragments through an interview without structure. writing-shape is the exploit step that organizes an existing pile into an article.
No. The raw material file is read-only to this skill, and it writes only to the article file you choose.