Manifest
Skill Kimotep/skills/manifest
Use this skill whenever the user wants to give form to the work in the current conversation — whether or not they know what they want. Triggers include "manifest this", "make this a document", "give this form", "ship this as a PDF", "make something I can share", "I want a one-pager", "turn this into something presentable", or simply "manifest". Two modes: (1) auto — the user says "manifest this" and the skill decides everything — format, structure, what to include, what to cut; (2) intent-guided — the user adds direction ("manifest this as slides for the board", "make a one-pager focused on the technical approach") and the skill executes against that intent. In both cases the skill reads context automatically. Output is a polished PDF — designed, never dense.From its SKILL.md
npx -y skills add Kimotep/skills --skill manifestAssembled from the repository path, not quoted from the project. Check it against their README if it does not work.
One thing to look at
- 0 stars0 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.
What its file declares
Copied from the file, not written here
The file declares its own license as MIT. That is the author’s claim about this one file, and it is not the same thing as the license GitHub reports for the repository, which is listed with the other numbers below.
SKILL.md
4.9 KB, ~1.0k tokens by cl100k_base, as published. Nobody here has run it
Manifest
Give form to the thinking. Whatever has happened in this conversation — a plan, an analysis, a decision, a comparison, a build — manifest it as something the user can hold, share, or hand off.
The PDF is the vehicle. The skill is editorial judgment: what matters, what to cut, how to structure it so the reader gets the point immediately.
The output should feel designed, not generated. Light, not dense. Every page breathes.
The skill runs through scripts/generate_pdf.py. Read it before generating.
Two modes
Auto — user says "manifest this" (or similar) with no further direction. The skill makes all the calls: what the document covers, what format it takes, what gets a full section vs. a callout vs. nothing. State the intent out loud before proceeding — "Manifesting a [type] doc covering [topics]." — then go immediately.
Intent-guided — user adds direction: "manifest this as slides for the board", "one-pager focused on the cost breakdown", "give this form as a data report". Honor the intent exactly. It overrides the skill's defaults on format, scope, and emphasis.
In both cases: no back-and-forth, no confirmation. Start unless the conversation is empty.
Phase 1: Read the room
Scan the conversation. Identify:
- What exists — what was produced, discussed, or decided?
- What it's for — external sharing? Internal reference? Pitch? Quick handoff? This shapes density and tone above everything else.
- User intent — did they specify format, audience, or angle? If yes, that wins. If no, choose based on the content.
- Style signal — did they say anything about look and feel? If not, default to
minimal. Userichwhen content is visual-heavy or design-forward.
Phase 2: Pick a format
| Format | When | Feel |
|---|---|---|
| report | Analysis, findings, summaries, plans | Sectioned, readable, editorial |
| slides | Things meant to be walked through, presented, pitched | One idea per page, visual |
| data | Comparisons, tables, metrics, charts | Numbers front, callouts for the headline |
| mixed | Most real work — a bit of everything | Flexible, section-by-section |
Default to mixed when in doubt. Default to report for short, focused output.
Phase 3: Build the spec
Assemble a Python dict. Every section should earn its place — don't pad.
spec = {
"title": "...",
"subtitle": "...", # optional — date, context, project name
"style": "minimal", # "minimal" | "rich"
"type": "mixed", # report | slides | data | mixed
"sections": [
{
"heading": "...",
"type": "text", # text | bullets | table | code | callout | diagram | chart
"content": "..." # string (text/code/diagram) | list (bullets) | list-of-dicts (table) | dict (chart)
}
]
}
Section types:
text— prose. Keep it tight.bullets—contentis a list of strings.table—contentis[{"Col": "val", ...}, ...]callout— editorial pull-quote. One insight only — no padding.code— code block.contentis a string.diagram— Mermaid source. Renders as a styled block.chart— bar chart.contentis{"labels": [...], "values": [...], "ylabel": "..."}
For slides: each section = one slide. One heading, one type, brief content.
Phase 4: Generate
python scripts/generate_pdf.py \
--spec-file /tmp/manifest_spec.json \
--output /path/to/output.pdf
Phase 5: Present
Share the file. One sentence on what it covers. Note anything omitted and why.
Editorial principles
- Lead with the point. Most important thing first.
- Never fill a page for the sake of it. Thin content = callout or bullet, not a padded paragraph.
- Callouts are for one thing. The single most important insight. If everything is a callout, nothing is.
- Every page breathes. Whitespace is design, not waste.
What ships with it: 3 files
17.6 KB alongside SKILL.md, 1 of them executable
scripts/
- generate_pdf.pyruns13.2 KB
- CHANGELOG.md1.5 KB
- README.md2.9 KB