Extract wisdom

Keep the useful ideas, insights, quotes, and recommendations.

Extract the most useful and surprising material from the supplied content.

Begin with SUMMARY in at most 25 words. Then use IDEAS for claims made in the source and INSIGHTS for carefully grounded implications. Add QUOTES with exact wording and speaker attribution, HABITS, FACTS REPORTED, REFERENCES, and RECOMMENDATIONS when the source supports them. End with ONE-SENTENCE TAKEAWAY in at most 15 words.

Use concise bullets, usually no more than 16 words, except exact quotations and titles. Select only the strongest items and remove duplicates across sections. Do not invent material to fill a quota. Separate reported claims from verified facts and your inferences from the speaker's statements. Omit empty sections. Preserve qualifications that change the meaning.

Example

Interview about incident-review habits

Before 92 words

In a team interview, Jo describes a weekly practice for reviewing support incidents. Before opening the incident log, Jo writes down the three incidents that come to mind. Jo then compares that list with the log to notice quieter recurring problems that memory missed. Jo says, "The loudest incident is not always the most common." Jo recommends trying the comparison for four weeks before changing the team's priorities. This is personal experience: the interview includes no measured results or controlled comparison. Jo does not claim that recall alone can rank incidents accurately.

After 82 words

SUMMARY Jo compares remembered incidents with logs to detect overlooked patterns; the practice has no measured results here. IDEAS - Jo says recall can miss quieter recurring problems. QUOTES - "The loudest incident is not always the most common." — Jo HABITS - Each week, list three remembered incidents before opening the log; then compare. RECOMMENDATIONS - Jo recommends four weeks of comparisons before changing team priorities. ONE-SENTENCE TAKEAWAY Use recall to expose attention gaps, not to rank incidents without the log.

Examples illustrate the method. They do not measure model output.

Files and sources

prompt.txt Complete prompt as plain text.
SKILL.md The same prompt with a skill header.
llms.txt (JSON) Prompt, metadata, example, and sources.
LICENSE MIT license text.

Adapted from Fabric · extract_wisdom. Source notes.