agentsclimarketplace

Interview synthesizer

Skill hamza-ali-shahjahan/hamzaish/_archive/v0/agents/idea/interview-synthesizer

Your Claude Code setup that builds, launches, and learns — across the whole product lifecycle. Point Claude Code at this and it runs your whole startup factory — and gets smarter every session. Works with Cursor, Codex & Windsurf too.

Install
npx -y skills add hamza-ali-shahjahan/hamzaish --skill interview-synthesizer

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

One thing to look at

  • 5 stars5 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

Synthesize a batch of customer interviews into evidence-for + evidence-against the hypothesis, with pattern surfacing and bias flagging.

SKILL.md

2.4 KB, as published. Nobody here has run it

Interview Synthesizer

When you activate

User has notes from 5+ interviews and asks: "synthesize these", "what did I learn from these calls?", "are we seeing PMF signals?"

What you produce

Saved to products/<name>/interviews/synthesis-<date>.md:

## Synthesis — interviews <date range> (N conversations)

### Evidence FOR the hypothesis
- <quote from interview> — (P3 — Sarah, Senior PM)
- <quote> — (P5 — ...)
- ...

### Evidence AGAINST the hypothesis
- <quote> — (P2)
- <quote> — (P7)
- ...

### Surprises (what we didn't expect)
- <observation> — surfaced by P4, P6
- ...

### Patterns
- N of N interviewees mentioned <specific behavior / phrase>
- <segmentation pattern>: <subgroup> behaves differently from <other subgroup>

### The strongest single quote
> "<verbatim>" — Person, Role

### Bias check
Compare lists:
- Evidence FOR: <count>
- Evidence AGAINST: <count>
- Ratio: <FOR/AGAINST>
- If FOR >> AGAINST, ask: is this real, or are we hearing what we want? Specifically: did our questions lead the witness? Did we ignore disconfirming signals?

### Recommendation
- CONTINUE: validation pattern is strong — refine and run 5 more
- PIVOT: the right problem nearby is <X>; rerun discovery
- KILL: evidence against is overwhelming — log learnings, move on
- INCONCLUSIVE: too noisy — change target profile or sharpen hypothesis

Protocol

  1. Read all the interview notes (paths the user provides, or products/<name>/interviews/raw/).
  2. Extract direct quotes — don't paraphrase. Cite by interviewee number/initial.
  3. Bucket into FOR / AGAINST / SURPRISE.
  4. Pattern-match across — note where multiple interviewees said the same thing in different words.
  5. Honestly check the FOR/AGAINST ratio for confirmation bias.
  6. Force a recommendation. No "we need more data" as a hedge if 5+ interviews already done.

Sources

  • knowledge-base/idea-stage/mom-test.md
  • knowledge-base/ai-native-2026/founders-playbook-distilled.md (re: bias check)

What you don't do

  • Don't paraphrase — quote.
  • Don't combine into "the average interviewee said X" — preserve disagreement.
  • Don't recommend "more data" unless you've named what specifically would resolve the question.

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.