mcpField
D
distill-response

distill-response

github.com/sjuneparkProductivity162 installsUpdated 3w ago
Repository

About distill-response

distill-response is a writing skill for turning selected source material into a concise explanation while preserving its conclusions, evidence, tradeoffs, uncertainty, and next steps. It provides guidance for both initial distillation and follow-up elaboration.

Key features

  1. 01Preserve conclusions and rationale
  2. 02Retain evidence and uncertainty
  3. 03Guide initial source selection
  4. 04Structure explanations by meaning
  5. 05Support follow-up elaboration

Use cases

  1. 01Condense a long response without losing key distinctions
  2. 02Create a standalone summary of selected material
  3. 03Expand a labeled point from an earlier summary

Install

Copy the command and run it in a terminal.

$npx skills add https://github.com/sjunepark/agent-scripts --skill distill-response

After installing, type/distill-responsein Claude Code to confirm it loaded.

Bundled filesagents/openai.yamlevals/evals.json

Details

From the project README

Distill Response

Preserve the source's decision spine: its conclusion, important relationships, consequential choices and rationale, evidence, tradeoffs, uncertainty, and next action where present. Compress repetition without losing distinctions that would change the reader's understanding or decision. Support every claim from the source and preserve its level of certainty.

Select the source

For an initial distillation, use the responses identified by the user, or the immediately preceding assistant response when none is identified. Read the complete selected source; honor any requested lens or length.

For elaboration, locate the most recent distilled map that established the requested label. An intervening explanation of a sibling item does not replace that map. For a subpart introduced during elaboration, use that explanation as the local source and retain the original map for orientation. Ask only when multiple maps or repeated labels leave the intended source materially ambiguous.

Deliver the explanation

For a distillation, lead with the outcome and organize around meaning rather than source order. Give distinct ideas stable descriptive labels when the reader may want to revisit them. Explain their relationships; use the smallest form that makes them clear, whether prose, a list, comparison table, or compact diagram. Simple linear material needs no forced labels or visual.

Keep an example, command, path, citation, exact value, or implementation detail only when it helps the reader understand, verify, decide, or act. Emit the distillation directly, with enough context to stand alone.

For elaboration, briefly reconnect the requested item to the established map, then resolve that uncertainty at the requested depth. Preserve existing labels, add stable sublabels when useful, and avoid repeating unrelated parts of the overview. Use examples or visuals when they clarify the requested point.

The result is complete when the reader can explain what matters, why it matters, the consequential choices and caveats, and what follows. An expanded item should still be easy to place within the original map.

Comments

Markdown supported