agentsclimarketplace

Synthesize interviews

Skill kimsanguine/signal-to-growth/skills/synthesize-interviews

Evidence-driven skills that turn customer signals into measurable growth.

Install
npx -y skills add kimsanguine/signal-to-growth --skill synthesize-interviews

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

2 things to look at

  • 12 days oldThe repository was created 12 days ago. New is not bad, but a brand new repository carrying a familiar-sounding name is the shape a typosquat arrives in, and there has been no time for anyone else to find a problem with it.
  • 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 author says it does

Copied from the file, not written here

Turn consented interview transcripts into source-linked evidence, themes, outliers, and counterevidence. Use when synthesizing customer interviews, VOC, JTBD research, 인터뷰 합성, or preparing evidence before a product or growth decision.

SKILL.md

2.5 KB, 457 tokens by cl100k_base, as published. Nobody here has run it

Synthesize Interviews

Create traceable synthesis without turning model-generated themes into approved customer truth.

Inputs

Require:

  • consented transcript or notes;
  • pseudonymous participant ID and role;
  • interview date and source file;
  • evidence and privacy policy;
  • the decision the synthesis should inform.

Reject summaries that contain no source material.

Workflow

  1. Preserve each transcript as a read-only source.
  2. Extract candidate quotes exactly and attach a file and line locator.
  3. Separate the quote from its interpretation.
  4. Assign stable EV-YYYYMMDD-NNN identifiers.
  5. Mark strength as awaiting_human_tag.
  6. Code evidence across participants without counting multiple quotes from one person as multiple people.
  7. Propose no more themes than the source can support.
  8. Record outliers, counterevidence, role differences, and missing segments.
  9. Ask a person to approve evidence strength and theme wording.
  10. Write the synthesis only after reference integrity passes.

Boundaries

  • Let the model extract candidate quotes, codes, and themes.
  • Use deterministic checks for exact source locators, IDs, distinct participants, schema, and references.
  • Require a person to approve evidence strength and any claim used in a decision.
  • Do not create a persona automatically.
  • Do not convert frequency into importance without context.
  • Do not use a fixed interview count as a universal gate.

Outputs

Create:

  • evidence.jsonl
  • theme-cards.md
  • counterevidence.md
  • synthesis-summary.md

Read references/output-contract.md before writing them.

Stop conditions

Stop when:

  • consent or source location is missing;
  • a quote cannot be found in the source;
  • participant identity cannot be pseudonymized;
  • a requested conclusion has no supporting evidence;
  • public output would expose restricted material.

Verification

Keep evidence.jsonl with the other run artifacts, then run:

python3 scripts/stg.py validate-artifacts artifacts/
python3 scripts/stg.py scan-privacy synthesis-summary.md

Every material claim must cite an evidence ID or be labeled as an inference, recommendation, or unknown.

Keep looking

Skills are one crate of 328,083. Ordering is by how many stacks a row turns up in, so the top of any crate is what has actually been picked rather than what has the most stars.