agentleFS
4444J994444J99/a-i--skills/skills/creative/natural-center-extraction/SKILL.md

natural-center-extraction

Skill17 starsChanged 5 months agoApache-2.0Archived repositoryView on GitHub

Apply narratological scoring to a text, transcript, or recording transcript — rating contiguous spans for pathos, logos, ethos, kairos, and density — then extract the single highest-scoring contiguous block as the work's Natural Center, with a scoring table and justification. Triggers on "find the natural center", "extract the most dramatic moment", "what's the strongest passage", or pulling the peak block from a long artifact for a short, excerpt, or trailer.

1 other project uses a version of this.

The file

We split this skill into the rules it gives an agent

8 rules. About 0 are common, 7 are distinctive, and 1 is very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈7

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 1

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Every sufficiently long artifact has a passage where its energies converge — the block a reader quotes, a trailer cuts to, an editor pulls.

    line 32
  2. Very specific:

    Contiguity is a hard constraint: no stitching, no elisions, no [...].

    line 58
  3. Distinctive:

    When composites tie (within ~5%):

    line 62
  4. Distinctive:

    Prefer the block that *survives decontextualization* — read it cold.

    line 63
  5. Distinctive:

    Prefer complete arcs (in-block setup and payoff) over blocks that borrow setup from outside.

    line 64
  6. Distinctive:

    Prefer the later block when scores tie exactly — late convergence usually means the whole work feeds it.

    line 65
  7. Distinctive:

    Always include the runner-up comparison — the justification is only credible against a named alternative.

    line 87
  8. Distinctive:

    One axis masquerading as five: If the winner leads every axis, re-check the scoring — axes are designed to disagree.

    line 97

≈ “Common” is an estimate: at least 1 in 5 of the 364,980 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Natural Center Extraction

Every sufficiently long artifact has a passage where its energies converge — the block a reader quotes, a trailer cuts to, an editor pulls. This skill finds that block by scoring rather than vibes, and returns it verbatim with the evidence for why it wins.

Why this exists

Excerpt selection is usually done by skimming, which favors openings and recency. Algorithmic scoring over all contiguous candidates surfaces centers that skimming misses — the buried turn in hour two of a transcript, the paragraph where argument and feeling finally coincide. The output feeds distribution work (shorts, pull-quotes, abstracts) with a defensible selection rather than a taste call.

The five axes

AxisWhat it measuresHigh-scoring signals
pathosfelt intensitystakes made personal; vulnerability; register shifts; concrete sensory detail
logosargumentative forcea claim completed in-block (setup → turn → consequence); evidence meeting assertion
ethosvoice authoritythe speaker most distinctly themselves; earned credibility on display; coinages dense
kairostimeliness/pivotthe moment the piece turns; before/after asymmetry; decision points; reversals
densitycompressionideas-per-sentence; no warm-up or wind-down inside the block; every line load-bearing

Each axis scored 0–10 per candidate block. Axes are scored independently then combined — a block must not be pre-selected because one axis shouts.

Workflow

1. Normalize and segment

  • Normalize the source to numbered units (paragraphs for prose; speaker-turns for transcripts; beats for scripts).
  • Generate candidate blocks: all contiguous runs whose length fits the target ±30% (sliding window). For long sources, pre-filter with a cheap density pass to a candidate set of 15–40 blocks before full scoring.

2. Score every candidate

  • Score each block on all five axes with one-line evidence per axis ("kairos 9: the 'and then I stopped' reversal lands mid-block").
  • Contiguity is a hard constraint: no stitching, no elisions, no [...]. The Natural Center is found, not assembled.
  • Compute the composite. Default: equal weights. Named intents shift weights — trailer/short → pathos+kairos ×1.5; abstract → logos+density ×1.5; bio/about → ethos ×1.5. State the weighting used.

3. Break ties structurally

When composites tie (within ~5%):

  • Prefer the block that survives decontextualization — read it cold; does it work without the surrounding pages?
  • Prefer complete arcs (in-block setup and payoff) over blocks that borrow setup from outside.
  • Prefer the later block when scores tie exactly — late convergence usually means the whole work feeds it.

4. Emit the result

# Natural Center — {source} ({date})
**Block:** units {n}–{m} ({word-count}w, target {t})  |  **Weighting:** {profile}

> {the block, verbatim, contiguous}

## Scoring table
| Block | pathos | logos | ethos | kairos | density | Σ |
|-------|--------|-------|-------|--------|---------|---|
| u12–u15 | 7 | 9 | 6 | 9 | 8 | 39 |  ← NATURAL CENTER
| u03–u06 | 8 | 5 | 7 | 4 | 6 | 30 |
| …top 5 candidates… |

## Justification
{one paragraph: why this block; which axes carried it; what the runner-up
lacked; per-axis evidence lines for the winner}
  • Always include the runner-up comparison — the justification is only credible against a named alternative.
  • Record the source span precisely so downstream cuts (video timestamps, page cites) can locate it.

5. Optional: multi-center mode

For serialized distribution (a thread, a cut-down series), re-run with the winner's span masked to find secondary centers. Label them C2, C3 — never promote a secondary to "the" center; the hierarchy is part of the finding.

Anti-patterns

  • Stitching a better block than the author wrote. Ellipses disqualify the result; contiguity is the constraint that keeps the method honest.
  • Scoring only the passages that caught your eye. The method's value is exhaustive candidate coverage; eye-catching passages are the skim bias this skill exists to defeat.
  • One axis masquerading as five. If the winner leads every axis, re-check the scoring — axes are designed to disagree; convergence is the finding, not the default.
  • Ignoring the target length. A brilliant 900-word block is a failed extraction when the ask was a 60-second short; fit is part of the score.
  • Editing the block "lightly" on output. Verbatim means verbatim; cleanup belongs to the downstream format pass, clearly marked as derivative.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.