Commands

Proof Points has three Slack commands and one app screen. Adding evidence is done in Slack: the app screen cannot take a source. The app is where you review, filter, run Scrutiny, and approve. Results from the Slack commands post in the channel where you ran them, and errors come back as a private message to you.

Add Evidence From a Source

/proof-extract <brand> <source text>

Name the brand first, then paste the source text. The first word after the command is always read as the brand, so do not start with the source itself. A private message confirms the request while the agent reads the source. Then a summary posts in the channel: how many new candidates were saved, marked unverified, and how many the source yielded in total. The candidates wait in Needs Review in the app.

If you run the command with no source text after the brand, you get a private message with the usage. Sources added this way are treated as your own work, so a point that names a client also gets a public version.

Verify and Score Candidates

/proof-scrutinize [brand]

Runs the Scrutiny pass for the brand and posts a summary with the number of points verified and the number marked as duplicates. Passing Scrutiny does not make a point usable. Approval in the app does.

Look Up the Library

/proof-points [brand] category=lead-generation min_conf=0.7

Posts a card of up to 15 matching points. Each entry shows the wording, its category, whether it passed Scrutiny, its confidence, a star if it is starred, and its attribution. The filters you can add are category=, min_conf=, verified=, starred=, and source_type=.

This command lists approved points only. It also defaults to points that passed Scrutiny, so add verified=false to see approved points that did not. Candidates waiting in Needs Review do not appear here.

Review in the App

Open the Proof Points screen in the app. From top to bottom:

• Counts: Needs Review, Total, Scrutiny Passed, Candidates, and Starred, for the selected brand.
• Filters: Category, Source Type (own or external), Verified (All, Verified Only, or Candidates Only), Starred, and Min Confidence from 0 to 1. Click Apply Filters, or Reset to clear them.
• Run Scrutiny: a button below the filters. It dedupes near-duplicates and scores survivors, then reports how many points were verified, culled, and demoted as duplicates. It does not make anything usable.
• Needs Review: each waiting point with an Approve button, a Reject button, and a field for an optional reason.
• Approved Library: the approved points, strongest first. Each card shows its type, category, Scrutiny Passed or Not Yet Scrutinized, Starred and Anonymized badges, the wording, any Public Version, the attribution, the source, a confidence percentage, and how many times it has been cited.

Approving or rejecting needs the permission to review proof points. The app has no control for adding a source or for starring a point.

Approve or Reject

• Approve: the point joins the Approved Library and becomes available to your drafting agents.
• Reject: the point is archived and will not be cited. The reason you type is saved with the point, and later extractions for the brand are told to avoid that claim and similar ones (the 20 most recent rejections are used). The app does not show the reason back to you.

Verify, Star, and Edit on the Operator Console

On the Operator Console, under Marketing, in Library, the Proof Points table lets operators toggle Verify or Star on a point and edit its wording and its public version. Changes are audited. Approval is not done there: it stays in the app's Needs Review queue.

Ask a Question in Slack

@Agenteous what proof points do we have on retention?

Mentioning @Agenteous gets a conversational answer. For the library itself, use the lookup command or the app.