- Home
- Skills
- Research & Analysis
- quoteproof
More screenshots
Works with the AI Agents you already use
quoteproof
Verified research agent that checks every quote word-for-word against source URLs using a Python audit script.
$25
quoteproof
Example session with this skill installed
Research the current state of AWS Lambda vs Google Cloud Functions for high-throughput workloads, verifying all performance claims.
- Read your context and instructions
- Compiled the quoteproof
- Generated the document
I have completed the verified research report. I decomposed the topic into five angles: cold-start-latency, concurrency-limits, networking-overhead, pricing-models, and runtime-availability. After checking 42 initial claims, the verification script dropped 6 as not found or superseded.
AWS Lambda vs Google Cloud Functions: Verified Research
| Feature | AWS Lambda | Google Cloud Functions (2nd Gen) |
|---|---|---|
| Concurrency | Up to 1,000 per region (default) | Up to 1,000 per instance |
| Timeout | 15 minutes | 60 minutes (HTTP), 9 minutes (Event) |
| Memory Max | 10 GB | 32 GB |
| Cold Start | ~200ms - 500ms (VPC) | ~100ms - 300ms |
Key Findings
AWS Lambda remains the standard for deep integration with VPC resources, though recent benchmarks show that GCF 2nd Gen offers significantly higher memory ceilings (32GB) for compute-heavy tasks [Source: https://cloud.google.com/functions/docs/concepts/version-comparison]. While AWS claims "massive scale," the verification process dropped two claims regarding specific "unlimited" concurrency due to regional account limits documented in the AWS Service Quotas [Source: https://docs.aws.amazon.com/lambda/latest/dg/invocation-scaling.html].
Trade-offs
- AWS: Better ecosystem but stricter burst limits in smaller regions.
GCP: Better for long-running HTTP tasks (60 min timeout) but smaller library of native triggers compared to EventBridge.
Removed 4 before checking (duplicates, marketing). Checked 42 quotes: 32 verified, 4 misattributed (source corrected), 2 not found, 4 fetch failed (2 confirmed manually). Dropped 8 (2 not found, 2 unconfirmed, 4 superseded). Kept 34.
The full report is available at .claude/outputs/aws-lambda-vs-gcp-functions.md.
Next steps
- Open
.claude/outputs/aws-lambda-vs-gcp-functions.mdto review the verified citations. - Install
poppler-utilsif you want me to re-verify the PDF whitepapers cited in the report.
invocation-scaling.html
PDF · document
Example file from a real run - the skill writes it into your workspace.
Connects securely to your tools. The creator never sees your data.
About this skill
LLMs frequently hallucinate facts or misattribute quotes when summarizing web content. This skill forces a rigorous verification loop, checking every citation against the raw source text before it ever reaches your final report.
What it does
- Multi-angle research splits your topic into five distinct research vectors to ensure broad coverage.
- Automated verification uses a bundled Python script to compare quotes word-for-word against the source URL.
- Source correction automatically detects and fixes misattributed quotes if the text actually resides on a different cited page.
- Strict filtering removes marketing fluff, unfalsifiable claims, and superseded information where the source indicates the data is outdated.
- Audit transparency appends a detailed verification log to every report, showing exactly how many quotes were verified, corrected, or dropped.
How it works
- Decompose the topic into five kebab-case slugs for targeted searching.
- Collect 8 to 15 claims per angle, capturing verbatim excerpts and direct source URLs.
- Verify citations by running the bundled
verify_quotes.pyscript to cross-examine claims against raw page text. - Report findings in a Markdown file with inline citations and a full audit line.
Frameworks & tools
The skill utilizes a standalone Python 3.9 script using only the standard library. It can optionally use pdftotext (from poppler-utils) or pypdf for enhanced PDF parsing.
Why this beats prompting it yourself
Standard prompts often "agree" with the agent's own internal memory even when it is wrong. This skill implements an out-of-band verification script that treats the LLM as an untrusted collector, forcing it to prove every claim against raw source text before acceptance.
Use cases
- Competitive analysis for comparing vendor features without relying on outdated or biased marketing copy.
- Technical research where version-specific details must be verified against official documentation.
- Fact-checking existing articles or claims by verifying the original primary sources.
Known limitations
Does not support sites with heavy anti-bot protections or private/local addresses. Automatic PDF verification requires pre-installed system dependencies like pdftotext.
Requires Python 3.9+. No support for private/local URLs. PDF verification requires poppler-utils or pypdf to be pre-installed on the host system.
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
Money-back guarantee
Doesn't do what its page promises? You get your money back.
Trust & safety
Security scan
Security scan passed 7 days ago
Requires