scroll →

1. 1Story header - Jira key, status and the org path. Nothing built yet.
2. 2Verdict first - who, what, where to fix.
3. 3Six reads fold into one row - queries in Details.
4. 4The broken link - Sales_Core gives no Read on the field.
5. 5What was not checked is listed.
6. 6Next step is a button on the card - with the users it touches.

scroll →

1. 1New story - no chat yet, the cards are the start.
2. 2Jira issue with its next step - one click starts the story.
3. 3Org alerts - Kestrel’s blocked login sits here too.
4. 4App queue - UAT done, Jira not updated yet.
5. 5Mail and chat stay off - labs only.
6. Empty state: No Jira connected - the first card lists open Changes instead.

scroll →

1. 1Org path - built in Dev-Alex, deployed to UAT, Production validated.
2. 2Plan card - what, where, risk, next step - the one warning sits here.
3. 3Three typed edit ops - each patches the version read first - 14 components read.
4. 4Visual diff - checkbox, layout section, text - whole files deploy.
5. 5Each number says where it comes from.
6. 6Way back is said before you approve.

scroll →

1. 1Approval is a card in the chat - its button is the only way to deploy.
2. 2Type client and class - a sandbox has the same org name.
3. 3The class is read from Salesforce, not from the typed text.
4. 4Tier ladder in the preview - this change is tier 2.
5. 5The approval holds for this exact plan only.

scroll →

1. 1One open card - the running deploy, Stop on it.
2. 2Stop also sits in the composer, and the story shows a running dot.
3. 3Timeline - done, now, estimated end.
4. 4Evidence only - each result as it lands.
5. 5Command targets the username - the org ID was checked before the run.

scroll →

1. 1The outcome, said in the chat - not “stop requested”.
2. 2Stop 09:22:05, canceled 09:22:41 - read back from Salesforce.
3. 3If Stop came too late, the preview says so and offers the snapshot.

scroll →

1. 1Result first - what was measured: Anna has Read.
2. 2Next steps sit on the card - restore or write the update.
3. 3What the assistant saw - names and counts.
4. 4Three proofs - diff, validation, screenshot with the row marked.

scroll →

1. 1Asked in plain words - the draft comes back as a card.
2. 2Two buttons - comment and status change are approved apart.
3. 3All three changes are in the text, each with its claim link.
4. 4Deploy time needs a confirm before it can go out.
5. 5Style checks against Alex’s past SFADM comments.
6. Empty state: No Jira connected - same draft, saved locally to copy out.

scroll →

1. 1Posted, with the exact comment link.
2. 2Read back - text, files and status, not “sent, hopefully”.
3. 3The preview shows the comment as Jira has it now - the story header says Done.

scroll →

1. 1Two paths - clients get workspaces, own org gets one.
2. 2Claude Code found and signed in - Codex is optional.
3. 3No agent - app opens in AI off, not on an error.
4. 4Sandbox for sure - Full or Partial known once Production is connected.
5. 5One org, no Jira, no Git - setup still ends in a first story.

scroll →

1. 1Verdict first - who must act.
2. 2Exact Salesforce error - for support tickets.
3. 3Why in two lines - with the help article.
4. 4Pack for their admin - four steps, nothing to guess.
5. 5Email goes to Outbox as a draft - nothing sent yet.

scroll →

1. 1Logins, rows, screenshots, voice - never leave.
2. 2Agent box - our tools only, no shell, no files, no tokens.
3. 3Our servers - Jira sign-in and updates, no org data.
4. 4Anthropic gets what your own Claude Code sends - rows only with your OK.
5. 5Per workspace - Kestrel runs with AI off.

scroll →

1. 1Ask in the story - the read shows which org it touched.
2. 2Read only - nothing in this panel writes.
3. 3Where used - 5 found, gaps said in Details.
4. Empty state: one org, no Jira - the same question works in a story without a ticket.

scroll →

1. 1Click Production in the path - overview card in the story.
2. 2Org card - class label from Salesforce.
3. 3This week - who, when, what.
4. 4Our own changes link to their receipt.
5. Empty state: one org - the path has one stage, the overview stays the same.

scroll →

1. 1Jira connected with Atlassian sign-in - one click to revoke.
2. 2Token fallback - opened here to show it - expiry date on screen.
3. 3GitHub App for the API, SSH key only for push.
4. 4Other workspace, other Jira - Kestrel runs Data Center.
5. 5No Jira, no Git - two "Connect" cards, nothing else breaks.

scroll →

1. 1One row per Change - what, why, where, state.
2. 2The path shows where each Change is now.
3. 3Kestrel’s Change waits on the blocked Production login.
4. 4Done Changes link their receipt.
5. Empty state: No Jira - the Why column shows your own note.

scroll →

1. 1Asked in plain words - no query to write.
2. 2Rows counted before anything changes.
3. 3Small and reversible - one click approves.
4. 4All or none - before-values saved first.
5. 5No Jira, one stage - Production is the work org here.
6. 6Rows go to the screen, not to the assistant.

scroll →

1. ① Guided card - ② highlighted display name - ③ before and after captures.

scroll →

1. ① Repair proposal - ② field error and exact patch - ③ check-only result.

scroll →

1. 1Docs sit in the sidebar footer - one set per workspace.
2. 2Story docs write themselves - what changed, why, receipts.
3. 3Object docs straight from the org - v1.5.
4. 4Imports keep the original - the agent reads the converted copy.
5. 5DOCX, PDF, XLSX/CSV, Markdown, images - drop or Import.
6. 6HTML by default - DOCX, PDF, Markdown out. Confluence in v1.5.
7. 7Every save is a version - same local Git as the changes.
8. 8Field mentions open the real field in the org.
9. 9Doc says 25 %, org says 20 % - flagged, not fixed.
10. 10Visual diff and receipt come in from the change.
11. Empty state: No stories yet - Import is the first button, nothing else.

scroll →

1. 1One request, one card - added and moved, in plain words.
2. 2Read first - what the agent looked at before the diff.
3. 3New value in green, where Setup will show it.
4. 4Who is hit - 0 records move, reports and flows Unknown.

scroll →

1. 1Ask in plain words - the card names the edit op.
2. 2Same layout twice - Discount leaves Other, lands in Pricing.
3. 3Who sees which layout stays the same - measured.

scroll →

1. 1Changed element in amber - same place in both versions.
2. 2New element in green - only in the draft.
3. 3Removed element - red, dashed, struck through.
4. 4The exact rule change, with how many records it hits.

scroll →

1. 1From the answer to one typed edit op - approve applies it in Dev-Alex.
2. 2One checkbox flips - Read on Discount, amber row.
3. 374 users get Read - measured in Production, Edit untouched.

scroll →

1. 1After - Discount in Pricing, help text on hover.
2. 2Sample data - built from describe, no real rows.
3. 3Approximation - custom components are placeholders.
4. 4Read first - what the preview is built from.

scroll →

1. ① Saved work and deploy proof - ② snapshot findings without a linked Change.

scroll →

Sales_Core.permissionset-meta.xml
@@ Opportunity.Discount__c · fieldPermissions @@
-    <readable>false</readable>
+    <readable>true</readable>

Base: b730ad1 · Change: 4c1e9a2
Normalizer: xml-safe-v1 · LF · UTF-8
Remaining component files appear above.

1. ① Reason and approver - ② three visual diffs - ③ prepare a restore Change.

scroll →

1. 1Look as a persona - profile + perm set group, or a real user like Anna Berger.
2. 2Diff list on top - each item points to its spot.
3. 3Headline: Prod picks a different page - the trail shows why.
4. 4Discount Reason exists only in UAT.
5. 5Margin % is on the Prod layout - FLS hides it.
6. 6Pricing moved - 2nd in UAT, 1st in Prod.
7. 7Apply Discount comes from Dynamic Actions - UAT page only.
8. 8Visibility rule shown, not guessed - sample Amount is 240k, so hidden.
9. 9Custom LWC - a box, not a render.
10. 10What was pulled - and what wasn't: record access needs a live query.
11. Empty state: One org - compare two personas in the same org instead.

scroll →

1. ① Tom’s request links to the Change - ② Teams is missing - ③ the suspicious email stays text.

scroll →

1. ① Snapshot and Git baseline - ② three unexplained changes - ③ excluded items.

scroll →

1. ① The save path - ② saved components - ③ the approval branch still needs checking.

scroll →

1. ① Counts are not a load result - ② matching uses a verified key - ③ rehearsal comes next.

scroll →

1. ① A guided check - ② the setting stays restricted - ③ both captures show the same value.

scroll →

1. ① Start in Dev-Alex - ② capture each result - ③ other workflows stay separate.