> ## Documentation Index
> Fetch the complete documentation index at: https://docs.libratech.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# Understanding Review results

> Risk badges, position Met counts, rule outcomes and citations: what each piece of a Review result tells you.

<video autoPlay muted loop playsInline src="https://mintcdn.com/libra-4206ec93/azhR9hqtFjs66ITC/assets/videos/features/review-results.webm?fit=max&auto=format&n=azhR9hqtFjs66ITC&q=85&s=a732277cf0b93b5f0a4ae3bef41496d4" className="rounded-xl border w-full" data-path="assets/videos/features/review-results.webm" />

When a Review finishes running, the result has the same shape every time. Once you can read one, you can read every Review, whether it has three Topics or thirty.

<div className="libra-buddy-callout">
  <img src="https://mintcdn.com/libra-4206ec93/wgVZXcGnHS4aJk9T/assets/images/libra-buddy/clipboard.png?fit=max&auto=format&n=wgVZXcGnHS4aJk9T&q=85&s=cde036cf23ea2d67cd50f1c295db549a" alt="Libra mascot with a clipboard" width="725" height="605" data-path="assets/images/libra-buddy/clipboard.png" />

  <div className="libra-buddy-callout-content">
    Edits you make to a Topic *can* travel back to the Playbook the Review came from — Libra asks you every time. Pick **Only this review** unless you really want the next matter to inherit the change.
  </div>
</div>

## How the result screen is laid out

The Review opens on three panes.

On the left, a sidebar with two tabs: **Results**, listing every Topic with its number and a coloured risk icon, and **Documents**, listing the files the Review ran against. Clicking a Topic in the list scrolls to it.

In the middle, the Topics themselves, numbered `01`, `02`, `03` and laid out one after another. Each one carries its risk badge, a **Reasoning** paragraph, and a **Positions** section.

On the right, the document viewer. It opens when you click a citation, and shows the cited page with the passage highlighted.

Above all three sits the header strip: the Review title, **Author**, **Updated**, an **Export** button, and either **Save as Playbook** or a **Saved as Playbook** marker.

### The levels of a result

```text theme={null}
Review
└── Topic            ← risk badge + Reasoning + Positions
    └── Position     ← Starting / Fallback / Not acceptable, each with a Met count
        └── Rule     ← outcome badge + reasoning + inline citations
```

Topics are laid out expanded, so you scroll rather than click to read them. Positions start collapsed: click a position row to open the rules underneath it.

## Topic risk badges

Every Topic that finished gets one badge.

| Badge           | Meaning                                                                                                                                                                                               |
| --------------- | ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
| **No risk**     | The contract meets the **Starting position** in full, and trips none of the **Not acceptable** rules.                                                                                                 |
| **Medium risk** | Anything in between — partial coverage of the Starting position, a **Fallback position** carrying the weight instead, or forbidden wording that shows up but is outweighed by the wording you wanted. |
| **High risk**   | The **Not acceptable position** matches at least as well as every other position on the Topic. The forbidden wording is as present as the wording you asked for.                                      |
| **Not found**   | Nothing in the uploaded documents was judgeable for this Topic. The Topic body says so: *"No matching content found in the uploaded documents for this topic."*                                       |

Two details of the rollup are worth knowing, because they change how you read a badge:

**No risk is a strict gate.** It needs both halves — the Starting position fully met *and* no rule met on any Not acceptable position. A contract that satisfies every rule you wanted but also contains a clause you forbade is not No risk. Fallback positions are deliberately left out of this gate; meeting a Fallback rather than the Starting position is a Medium-risk outcome, not a clean one.

**High risk is a comparison, not a threshold.** Libra scores each position by the proportion of its judged rules the contract meets, then asks whether a Not acceptable position scores at least as high as every other position on the Topic. So a Topic with no judgeable Not acceptable rules can never come out High risk: it is either No risk, if it clears the gate above, or Medium risk — however thin the coverage is.

While a Topic is still being evaluated it carries no badge at all — its card shows a placeholder shimmer, and its row in the left sidebar shows a spinner. A Topic that errored shows **Failed**, and you can re-run just that Topic.

<Tip>
  Scan the risk icons in the left sidebar before you read anything. A Review with twenty Topics is much less daunting when you start with the three red ones.
</Tip>

## Position Met counts

Each position row in a Topic shows how its rules came out, in the form `2/3 Met · 1 Not applicable`.

| Count                | What it means                                                                                                                                   |
| -------------------- | ----------------------------------------------------------------------------------------------------------------------------------------------- |
| **Met X/Y**          | Of the Y rules Libra could judge on this position, X had their condition hold in the contract.                                                  |
| **Not applicable N** | Rules the contract doesn't give Libra enough to judge. They are excluded from both X and Y and counted separately, so the figures still add up. |

**Met is literal, and it does not flip by position type.** It always counts the rules whose described condition holds — which means a high Met count is good news on a Starting or Fallback position and bad news on a Not acceptable one. A **Not acceptable position** reading `2/2 Met` means both forbidden clauses were found in the contract. A clean contract reads `0/2 Met` there.

That is deliberate: the neutral tally lives in the count, and the good/bad signal lives in the colours — the Topic risk badge above it and the per-rule badges inside it, which *do* change wording by position type.

A position whose rules are all **Not applicable** shows only the Not applicable count.

## Rule outcomes

Open a position and you see its rules, labelled **Rule 1**, **Rule 2** and so on, each with an outcome badge and a **Reasoning** paragraph.

The badge wording depends on the position the rule sits in, because "the condition holds" means the opposite thing on a Not acceptable position:

| Position type                                | Condition holds    | Condition doesn't hold   |
| -------------------------------------------- | ------------------ | ------------------------ |
| **Starting position**, **Fallback position** | **Met** (green)    | **Not met** (red)        |
| **Not acceptable position**                  | **Detected** (red) | **Not detected** (green) |

Two more badges appear on any position type: **Not applicable** in grey, meaning the contract doesn't give Libra enough to judge the rule either way, and **Pending** while the rule has not been evaluated yet.

Read the colour, not the word, if you're skimming: red is always the outcome you need to look at, whichever label it carries.

For example, on a Starting position:

> **Rule 1**: Weekly working time is between 40 and 50 hours.
> **Not met** — The working time is not clearly defined in the employment contract. Without a definition, it may exceed 50 hours per week.

<Warning>
  When a rule comes back red, always click through to the cited passage before you act on it. Most disagreements between Libra's reasoning and your own judgment turn out to be about what the contract actually says, and the citation settles that in one click.
</Warning>

## Citations

Citations are how Reviews stay verifiable, and they live inline in the reasoning text rather than in a list underneath it.

Each one renders as a small pill in the sentence it supports, carrying a file icon and the source document's name, truncated to fit. Where Libra can't resolve the file name, the pill falls back to a plain **Document** label. Anything Libra pulled from a legal database carries that research source's own icon and name instead. Where several citations back the same sentence, they collapse into one pill with a `+1`, `+2` counter.

Hover a pill and you get a card with the cited text; its footer offers **View in document** and, where the block has sources to list, **Citation list**, which opens every source cited in that reasoning block in a side panel. Click the pill itself and the document viewer opens alongside the result at the right page, with the passage highlighted.

## Auto mode Topics

A Topic run in Auto mode looks different, because it has no positions and rules to roll up.

It carries an **Auto mode** badge in its header, stacked with icons for the research sources it used, and the usual risk badge. Below that there is a **Reasoning** block and nothing else — no **Positions** section. The citations in that reasoning mix document pills with research-source badges, so you can trace both what the contract says and what Libra looked up.

A Review can mix Auto mode and rules-based Topics; both render in the same result with the same risk badges. See [Auto mode](/en/features/review/auto-mode).

## Marking a Topic as reviewed

A Review is a working document, not a one-shot output. Each finished Topic has a **Verify** button in its header — tooltip *"Mark as reviewed"*. Click it and the button becomes a **Reviewed** tag carrying your avatar; hovering the tag shows who verified the Topic and when.

Only the person who set it can clear it again, so a colleague picking up the Review can trust the tag. Verifying needs edit rights on the Review; view-only users see the tags but cannot set them.

There is one marking state, not two — a Topic is either verified or it isn't. If a Topic needs follow-up, leave it unverified and use your own matter notes.

## Changing a result

Rule outcomes themselves are not editable — you can't flip a **Not met** to **Met** in place. What you change is the Topic, and Libra re-evaluates it. Each Topic header carries three controls, all of which need edit rights:

<Steps>
  <Step title="Re-run evaluation">
    The circular-arrow button re-runs the Topic against the same documents, unchanged. Use it when a Topic failed, or when you think a retry will do better.
  </Step>

  <Step title="Edit topic">
    The pencil opens the Topic's positions and rules in an editor. Rewrite a rule that was too vague, add a missing one, drop one that doesn't belong. **Save** confirms with *"Topic saved, re-evaluating…"* and Libra re-runs that Topic on the new definition.
  </Step>

  <Step title="Delete topic">
    The bin removes the Topic from the Review.
  </Step>
</Steps>

Below the last Topic there is also an **Add topic** button, for a Topic the Playbook never covered. It obeys the same edit rights and the same busy gate.

Editing and deleting are blocked while the Review itself is still running and while any Topic is being re-evaluated; the buttons grey out with *"Wait for the running topic evaluation to finish before editing."* and *"…before deleting."*

### Whether the change reaches the Playbook

If the Review is attached to a Playbook you have edit rights on, saving an edit, an addition or a deletion opens a confirmation first. The title follows the action — **Update the original playbook too?**, **Add to the original playbook too?** or **Remove from the original playbook too?** — and each offers the same pair of choices, plus **Cancel** to back out.

| Choice                                                                                                                  | Effect                                                                                                            |
| ----------------------------------------------------------------------------------------------------------------------- | ----------------------------------------------------------------------------------------------------------------- |
| **Only this review**                                                                                                    | The change stays on this Review. The Playbook is untouched, and future Reviews start from the Playbook as it was. |
| **Update playbook & review** — or **Add to playbook & review** / **Remove from playbook & review**, matching the action | The change is written back to the Playbook as well, so the next matter inherits it.                               |

You won't see the prompt when the Review has no Playbook behind it, when you don't have edit rights on that Playbook, or when the Playbook has moved on to a newer version since the Review was created. In all three cases the change simply stays on the Review. The prompt is also skipped when you delete the last remaining Topic — Libra will not push a Playbook with no Topics back to the library.

<Note>
  If you find yourself making the same edit on Review after Review, that's the signal to fix the Playbook — either through **Update playbook & review** here, or on the [Playbook](/en/features/review/templates) itself.
</Note>

## Exporting a Review

The header has a single **Export** button. It downloads the finished Review as a Word document (`.docx`) — there is no format picker. The button is disabled until the run completes, with the tooltip *"Available once the review finishes processing"*.

Next to it sits the Review's Playbook link. A Review that came from a Playbook — or one you have since saved as a Playbook — shows a **Saved as Playbook** marker; hover it and Libra names the Playbook. A Review with no Playbook behind it shows a **Save as Playbook** button instead, which asks for a name and a description and turns the Review's Topics into a reusable Playbook for the next matter.

## Tips for reading Reviews efficiently

<AccordionGroup>
  <Accordion title="Triage by risk">
    Open the High-risk Topics first. Skim the Medium-risk ones. Trust the No-risk ones, but spot-check one or two for sanity — No risk is a strict gate, so it earns some trust.
  </Accordion>

  <Accordion title="Check the Met count against the position type">
    On a **Not acceptable position**, a high Met count is the bad outcome. If a position reads `2/2 Met` under a red Topic badge, that is two forbidden clauses found, not two boxes ticked.
  </Accordion>

  <Accordion title="Read the rule that surprises you">
    If an outcome surprises you, that is exactly the rule whose reasoning to read in full. It will either confirm Libra is right, or reveal a misread you can fix by sharpening the rule and re-running the Topic.
  </Accordion>

  <Accordion title="Treat citations as your audit trail">
    A Review is only as defensible as its citations. For client-facing or partner-facing work, click through every citation on the Topics you intend to negotiate.
  </Accordion>
</AccordionGroup>

## Next steps

<Columns cols={2}>
  <Card title="Reviews from chat" icon="https://mintcdn.com/libra-4206ec93/vIwxn8dmEk0jsP5x/assets/images/icons/chat.svg?fit=max&auto=format&n=vIwxn8dmEk0jsP5x&q=85&s=9a77dbd38fde763e8ac783fb8a68aef6" href="/en/features/review/from-chat" width="16" height="16" data-path="assets/images/icons/chat.svg">
    Use a finished Review as context in a follow-up conversation.
  </Card>

  <Card title="Playbooks" icon="https://mintcdn.com/libra-4206ec93/vIwxn8dmEk0jsP5x/assets/images/icons/review.svg?fit=max&auto=format&n=vIwxn8dmEk0jsP5x&q=85&s=034a75366c56b497bbe99cbb19f2f796" href="/en/features/review/templates" width="16" height="16" data-path="assets/images/icons/review.svg">
    Tune the Playbook if a result wasn't what you wanted.
  </Card>
</Columns>
