writing-shape
Writing, exploit——把原始素材塑造成文章,一段一段地推进。
Security Assessment
About writing-shape
writing-shape is a Chinese-language (translated) writing skill from the mattpocock-skills-zh-cn collection (in-progress path) that shapes raw material into a finished article, advancing paragraph by paragraph. It uses the explore/exploit framing from ideation: exploration is already done and the raw-material "pile" is fixed, so this skill is the exploit phase—commit to a structure and mine the pile to fill it. The word "exploit" here refers to that explore/exploit tradeoff, not to any security exploit. It is marked disable-model-invocation, so it is invoked deliberately.
The loop is: read the entire input pile first; establish prerequisites—what the reader knows on entry versus what must be grounded inside the article, treating each concept as something that must be grounded before a later block can depend on it; draft 2-3 candidate openings that each imply a different thesis and let the user choose or hybridize; then grow the article one block at a time, asking what the reader needs next and pulling material from the pile to answer, deciding deliberately whether each block should be prose, list, table, callout, quote, or code. It appends each agreed block to the article file immediately (re-reading from disk first, never blindly overwriting), pushes back on weak transitions, and names gaps when the pile lacks needed material. The raw-material file is treated as strictly read-only, and publishing/platform formatting is out of scope.
It targets writers turning notes, transcripts, or fragment piles into a coherent article with a single voice, who want a disciplined, grounding-aware, conversational drafting process rather than a one-shot generation.
FAQ
What does 'exploit' mean in this skill's description?
It refers to the explore/exploit tradeoff—the exploration is done and the raw-material pile is fixed, so this is the exploit phase of committing to a structure and mining the pile. It has nothing to do with security exploits.
What is 'grounding'?
Every concept must be grounded before a block can depend on it—either the reader brings it in as a prerequisite or an earlier block introduces it. The skill maintains a running list of what has been grounded so later blocks never reach for something undefined.
Does it edit my raw material file?
No. The raw-material pile is strictly read-only. The skill writes only to a separate article file, appending each agreed block and re-reading from disk before every write so it never blindly overwrites user edits.
How does it decide formatting for each block?
It discusses tradeoffs aloud—prose vs. list, inline vs. callout, table vs. repeated structure, quote vs. paraphrase, code block vs. inline code—so each format choice is deliberate and defensible.
Is it fully finished?
It lives under an in-progress path and sets disable-model-invocation, so it is invoked deliberately and should be treated as an evolving drafting workflow.
Install writing-shape
Quick Setup:
- Copy the skill folder to
.claude/skills/ - Claude will automatically detect and use the skill
Repository
vinvcn/mattpocock-skills-zh-cn