- Home
- Skills
- Research & Analysis
- research review
research review
Audit research memos and numerical claims against original sources using a two-pass verification ledger.
Free
Works with the AI Agents you already use
research review
Example session with this skill installed
Review this draft memo: 'Campaign B saw 18 sign-ups from 100 visitors, achieving an 18% conversion rate. This is 10% higher than Campaign A (8/100) and proves our new landing page is better.'
- Read your context and instructions
- Compiled the research review
I have completed the two-pass research review of the provided campaign memo. The audit identified one calculation error and one unsupported causal claim.
Research Review Ledger
| Version | Claim ID | Source Reference | Status | Correction/Revision |
|---|---|---|---|---|
| 1.0 | C-01 | S2: 18 sign-ups / 100 visitors | Supported | Claim of 18% conversion is numerically accurate. |
| 1.0 | C-02 | S1: 8/100 vs S2: 18/100 | Incorrect | Correction: This is a 10-percentage-point increase, but a 125% relative increase. The '10% higher' claim is mathematically misleading. |
| 1.0 | C-03 | Observed sign-up data | Unresolved | Status: Unsupported causal language. The data shows correlation, not proof of landing page efficacy without a controlled study method. |
Revised Memo Draft
Campaign B recorded 18 sign-ups from 100 visitors, resulting in an 18% conversion rate. This represents a 10-percentage-point increase over Campaign A (8/100). While the observed sign-up volume is higher, the current data set does not isolate the landing page as the sole causal factor.
Next steps
- Update the memo with the percentage-point distinction to avoid misrepresenting growth.
- Remove 'proves' and 'better' to reflect the observational nature of the data.
- Provide the study methodology if a significance test was performed to resolve C-03.
Connects securely to your tools. The creator never sees your data.
About this skill
Research memos and numerical recommendations often fail due to miscalculated rates, shifted time windows, or causal leaps not present in the source data. This skill provides a structured evidence-checking layer to verify claims against original sources before you rely on the findings for decisions. It moves beyond simple proofreading to audit the logic and math underlying your data.
What it does
- Evidence mapping connects specific draft claims to original source passages using stable claim IDs.
- Numerical audit recomputes rates, totals, and populations to catch calculation errors.
- Logic validation flags unsupported causal language and recommendations that exceed the data scope.
- Conflict resolution escalates unresolved disagreements between data and drafts rather than looping.
How it works
- Ingest the decision, original sources, and draft to assign claim IDs and identify missing data.
- Researcher pass generates a claim ledger containing source passages and re-calculated figures.
- Reviewer pass audits the ledger against original inputs to verify time windows and populations.
- Correction returns a consolidated change log with revised memo text and unresolved flags.
Frameworks & tools
This skill is framework-agnostic and functions within any LLM environment capable of multi-pass reasoning or multi-agent orchestration.
Why this beats prompting it yourself
Standard prompts often mirror the assistant's own confidence or skip deep numerical verification. This skill enforces a structured two-pass ledger system that separates claim extraction from verification, ensuring calculations are re-run rather than just skimmed.
Use cases
- Auditing internal research memos before executive presentation.
- Verifying conversion rate calculations and A/B test summaries.
- Checking technical whitepapers against cited source data.
- Reviewing financial summaries for consistent time-window reporting.
Known limitations
This skill does not authorize external data disclosure or experiment rollout. It ignores spelling-only edits and requires access to the original raw sources to function.
Find more skills like this on the Claude Skills Marketplace, the Codex Skills Marketplace and the Gemini Skills Marketplace, or browse Claude Skills for Research & Analysis.
Does not handle spelling-only edits. Requires original sources; missing sources result in unresolved flags. Not an independent validation if run in a single-agent conversation.
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 buying a skill.
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
1 use
Times used to date
Free to use
No account required to browse
Trust & safety
Security scan
Security scan passed 24 days ago
- Free with an Agensi account
Permissions
No special permissions declared or detected