Works with the AI Agents you already use
Claude Code
Cursor
Codex CLI
Muse
Dots+22 moreClaude CodeCursorCodex CLIMuseDots
OpenClaw
Manus
Grok
Grok Bot+18 agentsdrm audit
Audit text for evidence quality, logical contradictions, and unsupported certainty using a rigorous labeling system.
Free
drm audit
Example session with this skill installed
Audit the following conclusion from a technical summary: "Our new caching layer reduces latency by 40% in all scenarios, proving it is the most efficient solution on the market." Context: The data shows a 40% drop in p99 latency for read-heavy workloads, but write-heavy workloads saw a 5% increase. No other solutions were tested.
- Read your context and instructions
- Compiled the drm audit
I have completed the DRM Audit of the provided technical summary regarding the caching layer performance claims.
DRM Audit Report
1. Scope
- Material reviewed: Technical summary of caching layer performance.
- Audit question: Does the evidence support the claim of 40% reduction in all scenarios and market superiority?
- Limitations: Review is restricted to the provided summary only.
2. Executive summary
- The claim of 40% reduction "in all scenarios" is contradicted by the provided data.
- The claim of being the "most efficient solution on the market" is an unverified hypothesis.
- Data confirms significant improvement specifically for read-heavy workloads.
3. Evidence ledger
| ID | Claim / excerpt | Label | Support | Reason | Next check |
|---|---|---|---|---|---|
| C1 | "reduces latency by 40% in all scenarios" | [CONTRADICTION] | Low | Text notes a 5% increase in write-heavy workloads. | Update claim to specify workload types. |
| C2 | "proving it is the most efficient solution" | [UNVERIFIED] | Not assessable | No comparative data for other market solutions was provided. | Benchmarking against competitors. |
| C3 | "40% drop in p99 latency for read-heavy" | [SUPPORTED] | High | Directly stated as a measurement in the provided data. | None required. |
4. Contradictions and position shifts
| Pair | Excerpt A | Excerpt B | Classification | Why |
|---|---|---|---|---|
| P1 | "reduces latency by 40% in all scenarios" | "write-heavy workloads saw a 5% increase" | Contradiction | "All scenarios" is a universal set that includes write-heavy tasks. |
5. Unsupported certainty
- Quote: "proving it is the most efficient solution on the market"
- Overreach: Claiming market-wide superiority without testing alternatives.
Defensible wording: "showing significant p99 improvements for read-heavy workloads compared to our previous baseline."
8. Final assessment
The text establishes a significant performance gain for read-heavy workloads (40% p99 reduction) but fails to support its universal claims. The conclusion of market superiority is currently a hypothesis.
Next steps
- Remove the "all scenarios" phrasing to reflect the 5% increase in write-heavy tasks.
- Strike the market superiority claim until competitive benchmarking is performed.
Connects securely to your tools. The creator never sees your data.
About this skill
Verify the integrity of technical reports, AI responses, and research papers without the fluff. This skill provides a rigorous framework for auditing supplied text for evidence quality, unsupported certainty, and internal contradictions. It forces a clear separation between what is directly observable in the text and what is merely an interpretation or hypothesis.
What it does
- Evidence mapping tags claims as supported, interpreted, or unverified based strictly on the provided material.
- Contradiction detection identifies incompatible claims by comparing definitions, scope, and context.
- Support leveling assigns high, medium, or low support ratings with specific rationales for every material claim.
- Inference auditing flags conclusions that are stronger than the underlying evidence or cite motives without proof.
- Discriminating tests proposes specific, practical checks or controls to resolve remaining uncertainties.
How it works
- Materiality gate performs an initial pass to determine if the text contains material issues before deep auditing.
- Evidence ledgering builds a structured table of claims, exact quotes, and support levels.
- Certainty review audits for overreach, such as using "always" or "proves" when the data only shows correlation.
- Final reporting delivers a structured audit report with clear distinctions between facts and hypotheses.
Frameworks & tools
This skill is framework-agnostic and works with any text-based input including AI logs, technical specifications, and research summaries.
Why this beats prompting it yourself
Generic prompts often hallucinate flaws to appear helpful or miss subtle shifts in scope that mask contradictions. This skill uses a strict labeling system and a materiality gate to ensure findings are grounded in exact quotes rather than vibes.
Use cases
- AI response auditing to check if a model's output is actually supported by its training data or RAG context.
- Technical proposal review to identify gaps where a project plan assumes outcomes without specific milestones.
- Research validation to separate a paper's raw findings from the author's subjective interpretations.
Known limitations
This skill only audits the text provided and cannot perform external web browsing or fact-checking. It does not prove truth in the real world, only internal consistency and support within the text.
How to use
Skills on Agensi run inside your AI agent. Copy the prompt, paste it into your agent, and it connects to Agensi and installs this skill for you. If a payment is needed, your agent shows it to you first.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill passes our trust checks below.
Security scanned
Passed our 8-point scan before listing
2 installs
Downloaded by developers to date
Free forever
No account required to browse
Trust & safety
Security scanned
Verified clean 5 days ago
- Free to download with an account