More screenshots
Works with the AI Agents you already use
Trustworthy Insights: Customer Themes Without Fake Quotes
AI summaries of interviews and surveys invent quotes, stitch quotes together, and state thin findings with full confidence. Your agent checks every quote against the source and counts people, not mentions.
Free
Trustworthy Insights: Customer Themes Without Fake Quotes
Example session with this skill installed
Here are 14 interview transcripts with churned customers. What are the main reasons they left? We are deciding whether to rebuild onboarding.
- Read your context and instructions
- Compiled the trustworthy insights
The agent opens with the sample: 14 people across 11 accounts, no enterprise accounts in the set. It reports three findings, each with a count (for example, 6 of 14 people across 6 accounts described setup taking more than a week), a confidence level with its reason, what it means for the onboarding decision, and verbatim quotes cited by participant and timestamp. One finding is marked LOW because it comes from 2 people at one account. It lists a conflict: a participant who said onboarding was fine but described abandoning setup twice.
Connects securely to your tools. The creator never sees your data.
About this skill
About this skill
Ask an agent to find themes in customer interviews and it will return clean, quotable findings. Some of the quotes do not exist in the transcripts, some are two sentences from different places joined together, the themes are generic enough to fit any product, and a point one person made ten times reads as strongly as one ten people made.
This skill adds checks at three points. Before reading, the agent writes down the decision the analysis feeds and how participants were picked. While extracting, a quote is one continuous stretch of one person's words, hedges stay in, and every quote carries a participant ID and location. Before reporting, the agent searches the source for each quote and marks it found, paraphrased, or missing, deletes the missing ones, checks each participant's stated preferences against their described behavior, counts people instead of mentions, and marks any finding from fewer than 3 people or one account as low confidence.
The proof
There is no eval score for this skill. Its proof is a check you can audit: every quote in the output is searched in the source and labeled FOUND, PARAPHRASE, or MISSING before the report goes out, and every finding states how many people and accounts support it. The output opens with one line giving the sample size and what is missing, so a reader who stops there knows what the analysis can support. The method credits Caitlin Sullivan's article "How to do AI analysis you can actually trust" in Lenny's Newsletter; the specific checks are the author's own.
What you get
One SKILL.md with the before, during, and after checks and a fixed output template: sample line, finding, confidence with its reason, decision impact, verbatim evidence with citations, and conflicting quotes.
How to use
- Add the skill folder to your agent's skills directory (for Claude Code,
~/.claude/skills/trustworthy-insights/). - Give the agent your transcripts, survey export, or tickets as text, and tell it the decision the analysis is for.
- Read the sample line first, then each finding's confidence and the quotes behind it.
Source and license
Free to use and MIT licensed. The full source, install steps, and issue tracker are on GitHub at https://github.com/robertmnyborg/analyst-rigor-skills.
Works only on text it has been given and can only verify quotes against that text. It does not transcribe audio or pull from survey tools. It does not do statistics on quantitative survey results. Confidence levels are rule-based (people and account counts), not statistical tests.
How to use
- 1
Connect your agent to the Agensi MCP
Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup
- 2
Sign in to Agensi
Your agent shows a sign-in link. Sign in and click Allow.
- 3
Ask for what you need
Your agent loads this skill and uses it on your task. It always asks before any payment.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill goes through the trust checks below.
Security scanned
Checked by our 8-point scan before listing
Fresh listing
Recently published to Agensi
Free forever
No account required to browse
Trust & safety
Security scan
Security scan passed 8 days ago
- Free with an Agensi account
Requires
Permissions
No special permissions declared or detected