Review startup claim evidence with AI
Trace important startup claims to supplied evidence, preserve conflicting sources and prepare a register of gaps and follow-up questions.
Review startup claim evidence with AI by separating what a document asserts from what the supplied material supports. The review-startup-claim-evidence skill connects material statements to exact locations and records their evidence state in a traceable register.
If two documents use different definitions of a customer, both locators remain visible alongside a clarifying question. Missing evidence means unverified, not false. The readable table and machine-readable register prepare a bounded human review; they do not produce a company score, valuation or investment recommendation.
What it does
- Build a claim-to-evidence register
- Preserve precise page, slide or row locators
- Compare dates, units and definitions across documents
- Distinguish supported claims from unverified ones
- Prepare questions for missing or conflicting support
- Explain what a resolved question would change
Install and use
- Load the full folder in a file-capable agent and provide Node for structural validation. Select the documents, date boundary and review purpose. No company-scoring service account is required.
- Ask: “Compare the customer and revenue claims in these two files. Preserve definitions, periods and currencies, and write questions for discrepancies.”
- Check that each row contains one testable claim and real source locations. Missing support must stay missing rather than being filled from memory or invented sources.
- Review the readable table and structured file alongside the local validation result. Retain conflicts, unreviewed materials and limits; structural checks cannot authenticate documents or establish evidence truth.
FAQ
Source
Made by Playful Pixels Oy. Original skill: review-startup-claim-evidence, published on GitHub under the MIT licence.
View on GitHub