Try the beta · Fictional examples

Try the writing. Check the setup.

Start with a short writing test. Add scoring only if you want to test the connection too. Either result is useful feedback.

No real resume, email connection or application is needed. Use your own copy of Stitch Your Pitch in the Grok Bot desktop app and record your operating system and app version. Keep mobile reports separate. This round covers Grok Bot; ChatGPT setup is separate.

Open the Bot Connection instructions

Optional: change one requirement · Copy the report fields

Test A: two messages, no tools

Add the template and open your copy first. Then copy each message below into the Bot. The same instructions apply if you already have a scoring connection.

Message 1: draft from evidence

This is a fictional writing test. Do not call tools, search for jobs, read email, create files or submit anything. Do not produce a numerical fit score.

Resume: Alex Sample, Sales Associate at Example Retail. Two years of retail sales experience. Used Salesforce to track customer follow-ups. Prepared weekly sales reports in Excel.

Job: Sales Associate at Example Store. Required: retail sales experience and Excel. Preferred: Salesforce and SQL. Responsibilities: customer follow-ups and weekly sales reporting.

Show a requirement, exact evidence and gap table. Then draft a concise cover letter using only these facts. Keep the evidence notes outside the letter.

Message 2: revise the voice

Revise the letter to sound warmer and more natural. Avoid em dashes, contractions and generic claims such as "perfect fit" or "proven track record". Keep the same facts, do not add SQL experience or invented enthusiasm, and show which sentences changed outside the revised letter. Continue without tools or a numerical score.

What to check

Sales experience, Salesforce follow-ups and Excel reporting stay tied to the supplied facts.
SQL stays a preferred gap. No new qualification, metric, employer or motivation appears.
The revised letter follows the requested style. Evidence notes stay outside the letter.
No tool call, file creation, numerical score or application action occurs.

Quote any fictional sentence that feels generic or inaccurate and explain why. Good tone does not excuse an invented fact. This test makes no AI-detector or human-authorship promise.

Public template content is 0.4.1 beta. You can request wording changes in conversation. The 0.4.2 writing-settings workflow is still a candidate and is not part of this published testing round.

Optional Test A2: change one requirement

Keep the resume and writing preferences fixed. Compare two postings that differ in one line: Excel becomes SQL. This tests tailoring separately from scoring.

Copy this prompt after Test A. One conversation is enough for a first pass; record the posting order and whether you used the same or separate conversations.

This is a fictional comparison. Use only the resume and postings below. Do not call tools, calculate numerical scores, search, read email, create files or apply. Use plain professional tone, concise length, no em dashes and no contractions for both postings.

Resume: Alex Sample. Sales Associate at Example Retail. Two years of retail sales experience. Used Salesforce to track customer follow-ups. Prepared weekly sales reports in Excel.

Job A:
Sales Associate at Example Store.
Required: retail sales experience.
Required reporting tool: Excel.
Preferred: Salesforce.
Responsibilities: customer follow-ups and weekly sales reporting.

Job B: identical to Job A, replacing only "Required reporting tool: Excel." with "Required reporting tool: SQL."

For each job, produce a requirement/evidence/gap table, a tailored resume summary with two bullets and a short cover letter. Review each resume and each letter separately against the unchanged source. Keep factual review notes outside the drafts. Explain which changes in wording or emphasis follow from the changed requirement, including any justified unchanged text. Keep missing qualifications visible in the review. Never change a fact to match the posting.

Job A: supported Excel reporting gets relevant emphasis.
Job B: SQL is a required gap. Excel experience remains truthful and never becomes SQL experience.
Each resume and each letter has its own factual review. Accuracy in one does not certify another.
Unchanged relevant wording is valid. Count factual errors and useful changes, not just changed words.

The gap belongs in the review; it does not need to become an unnecessary self-rejection in the letter. Never claim the missing qualification is held. Report specific sentences separately for the resume and letter. One trial does not establish a general effect.

Test B: template and scoring setup

Use these step labels in your report. The template adds seven skills to your own Bot copy. The optional scoring connection is added separately and uses your own OAuth sign-in. Stop at a failed or uncertain step.

S0 · Preview
Open the public share link and confirm the template preview loads.
S1 · Add your copy
Add the template and open the resulting copy in Grok Bot desktop. If it is already installed, record that instead of adding another copy.
S2 · Check seven skills
Check Context for seven skills. Record missing names or truncated content if visible. A preview alone does not establish an import.
S3 · First response
Send Test A message 1 and check for a useful response without tools.
S4 · Add the connection
Follow the current setup instructions to add the custom public scoring connection.
S5 · Your sign-in
Complete your own Auth0 sign-in and return to Grok. Confirm only scoring and normal renewal access.
S6 · Discover the tool
Confirm score_job appears. A connected badge alone is insufficient.
S7 · One fictional score
Send the message below once. Record the actual result status and request ID, when supplied.

For this one request, lift the earlier no-tool instruction only for a single score_job call using the fictional resume and job from Test A message 1. Validate the evidence-linked inputs, call score_job once and show the actual result, uncertainty notes and request ID if supplied. Do not call other tools or retry automatically. If the call fails or its outcome is uncertain, stop and report that status without inventing a score.

Scores are heuristics, not an employer's ATS result or a hiring prediction. No fixed score is promised for model-extracted inputs. A separate frozen-input test has an expected score of 90; that is not the expected result for this conversation.

If a skill is missing or setup fails, keep the copy and report the step. Do not delete and reimport it, change clients mid-report, copy publisher credentials, broaden permissions or retry an uncertain scoring request automatically. Do not test quota exhaustion. Renewal and reopening tests are separate follow-ups.

Send a useful report

Share the relevant fields through the support channel or the beta discussion where you found this guide. Leave out real resumes, account emails, credentials, authorization codes, tokens, raw identities and private screenshots.

Test: A writing / A2 one requirement / B setup
Client and operating system:
App version:
Displayed template content version, if visible:
First install or existing copy:
Connection already present: Yes / No / Unknown
Last successful turn or setup step:
Next turn or setup step attempted:
Expected result:
Actual result or sanitized error:
Approximate time and time zone:
Request ID, if supplied:
Specific fictional sentence that felt generic or inaccurate:
A2 only: posting order; same or separate conversations:
Did you retry or switch clients? Describe what happened:

For example: "S5 succeeded; S6 attempted; score_job did not appear." That identifies tool discovery, not login, as the failing step. An uncertain S7 outcome stays unknown. An existing connected account does not establish fresh-user setup.

Documented checkpoint · November 5, 2026

Observed: Three successful fictional scoring calls from the owner's account. Auth0 records confirm successful renewal for that account.
Still open: Independent users installing their own copies, completing sign-in, receiving independent quotas and returning after expiry with client reopening.
What this means: Owner testing does not establish the new-user experience. This dated checkpoint is not live monitoring, an availability promise or a hiring prediction.

Public template content is 0.4.1 beta. App metadata can display an older date. The template includes seven skills; the optional custom connection remains a separate setup step.