agentleFS
RylaispiritRylaispirit/rylai-codex-hermes-skills/skills/summarization/SKILL.md

summarization

Skill56 starsChanged 3 months agoMITView on GitHub

Create accurate summaries of documents, articles, meetings, research, code changes, or multiple sources as briefs, key points, actions, or catch-up notes.

The file

Show its rules and text

We split this skill into the rules it gives an agent

20 rules. About 0 are common, 20 are distinctive, and 0 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈20

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 0

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Use tools available in the active runtime to read the actual source material.

    line 24
  2. Distinctive:

    Do not claim full coverage of inaccessible, truncated, or unsupported input.

    line 25
  3. Distinctive:

    Verify extraction quality when the source is a scan, recording, or complex file.

    line 26
  4. Distinctive:

    When the request is underspecified, choose a practical format and state the compression level in one short line.

    line 39
  5. Distinctive:

    Keep what the target reader needs to decide, act, understand, or verify.

    line 54
  6. Distinctive:

    Remove repetition, scene-setting, and low-impact detail unless requested.

    line 55
  7. Distinctive:

    Put the bottom line first.

    line 58
  8. Distinctive:

    Check names, figures, dates, units, ownership, and negation.

    line 63
  9. Distinctive:

    Use for reports, proposals, and decision material:

    line 71
  10. Distinctive:

    Use for specifications, architecture, incidents, and code:

    line 81
  11. Distinctive:

    Use for notes, chats, and email threads:

    line 92
  12. Distinctive:

    Do not assign an owner or deadline that was not explicitly agreed.

    line 100
  13. Distinctive:

    Use for papers and analytical material:

    line 104
  14. Distinctive:

    Keep correlation, causation, statistical confidence, and author interpretation distinct.

    line 112
  15. Distinctive:

    Read all available sources before drafting.

    line 117
  16. Distinctive:

    Avoid presenting repeated claims as independent confirmation.

    line 121
  17. Distinctive:

    Do not invent context, intent, consensus, or causality.

    line 133
  18. Distinctive:

    Use direct quotes only when wording itself matters and the source is available.

    line 135
  19. Distinctive:

    Keep recommendations separate from the summary unless the user requests them.

    line 138
  20. Distinctive:

    All critical figures, decisions, and caveats match the source.

    line 148

≈ “Common” is an estimate: at least 1 in 5 of the 350,495 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

Summarization

Reduce reading time while preserving the source's meaning, evidence, decisions, uncertainty, and useful detail.

Runtime Compatibility

  • Resolve bundled files relative to this skill directory.
  • Use tools available in the active runtime to read the actual source material.
  • Do not claim full coverage of inaccessible, truncated, or unsupported input.
  • Verify extraction quality when the source is a scan, recording, or complex file.

Intake

Infer these settings from the request and source:

  • audience;
  • purpose;
  • output length;
  • required format;
  • facts that must survive compression;
  • whether attribution to speakers or documents matters.

When the request is underspecified, choose a practical format and state the compression level in one short line.

Workflow

  1. Establish source boundaries
    • Identify every file, message range, URL, transcript, or diff included.
    • Note missing pages, failed extraction, inaudible segments, or truncation.
  2. Build a source map
    • Record the central claim or purpose.
    • Capture decisions, evidence, dates, quantities, people, risks, and open items.
    • Separate facts, opinions, proposals, and unresolved statements.
  3. Rank information
    • Keep what the target reader needs to decide, act, understand, or verify.
    • Remove repetition, scene-setting, and low-impact detail unless requested.
  4. Write the summary
    • Put the bottom line first.
    • Group related points rather than following source order mechanically.
    • Preserve qualifications that change the meaning of a claim.
  5. Audit against the source
    • Check names, figures, dates, units, ownership, and negation.
    • Confirm that no new conclusion was introduced.
    • Flag contradictions instead of silently resolving them.

Format Selection

Executive Brief

Use for reports, proposals, and decision material:

  • bottom line;
  • strongest evidence;
  • risks or uncertainty;
  • required decision;
  • next actions.

Technical Summary

Use for specifications, architecture, incidents, and code:

  • purpose and scope;
  • behavior or design;
  • interfaces and dependencies;
  • important decisions;
  • trade-offs and limitations;
  • breaking changes or follow-up work.

Meeting Or Conversation

Use for notes, chats, and email threads:

  • decisions;
  • action items with owner and due date when stated;
  • key reasoning;
  • disagreements;
  • unresolved questions.

Do not assign an owner or deadline that was not explicitly agreed.

Research Digest

Use for papers and analytical material:

  • research question;
  • method and sample;
  • main findings;
  • limitations;
  • implications.

Keep correlation, causation, statistical confidence, and author interpretation distinct.

Multi-Source Synthesis

  • Read all available sources before drafting.
  • Organize by theme or question.
  • Attribute disputed or source-specific claims.
  • Surface conflicts in dates, measurements, definitions, or conclusions.
  • Avoid presenting repeated claims as independent confirmation.

Compression Levels

  • Snapshot: one paragraph or a few bullets.
  • Brief: bottom line plus essential evidence and actions.
  • Detailed: structured coverage of all major sections and caveats.

Scale by information density and user need, not page count alone.

Output Rules

  • Do not invent context, intent, consensus, or causality.
  • Preserve exact numbers and dates when they affect meaning.
  • Use direct quotes only when wording itself matters and the source is available.
  • Mark uncertainty with precise language.
  • Distinguish source omissions from your own inability to access material.
  • Keep recommendations separate from the summary unless the user requests them.

Optional Templates

Load summary-templates.md when a reusable output skeleton will save time. Adapt the template to the source instead of filling empty sections mechanically.

Final Check

  • The opening states the most important conclusion or purpose.
  • All critical figures, decisions, and caveats match the source.
  • The summary is useful without pretending to replace unavailable evidence.
  • The length and terminology fit the intended reader.

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.