Where scores appear

Open Agentic Foundation in the sidebar to see every connected repo in a list. The workspace header leads with N of M repos measured. The % next to it is Agentic readiness. The section title stays Overall Agentic Foundation.

That % is the mean overall repo-agent maturity from Agentic Foundation scans (0–100). It includes unscanned repositories as 0%. If the workspace has repos but none have been scanned yet, you see 0% and the line includes unscanned as 0%, with no band chip. If there are no repositories, the page shows No repositories analyzed yet.

Agentic readiness measures repo-agent maturity from scans; the engineering delivery score measures live delivery signals from connected integrations.

Use Show unscanned on the table when you want only repos that have not been scanned yet.

Each row shows Overall (0–100) and a color band after a real scan. Unscanned rows stay muted and show Not scanned instead of 0%. Click a repo name to open its detail view, where the same number sits at the top with a per-dimension breakdown underneath. Under the repository name, the path (such as org/repo) opens that repository in your source control host. Next to Run analysis, Open in… lets you choose Cursor (desktop), Cursor Web, your default IDE, or Claude Code. Nothing opens by itself; you click. Under the Dimension Breakdown title you will see a short section description, then a note that links to Code Insights for complexity and hotspots from the same scan. That is separate from readiness Fix-it items.

Workspace score over time

The Agentic Foundation overview (/agentic-foundation) shows one workspace progression chart. It is not a nine-chart grid, and it is not on a repository detail page.

The chart appears only after the workspace has two or more completed scans. Until then the hint is Trend after two scans. and the card stays hidden. A workspace with no repositories still shows No repositories analyzed yet and no chart.

First paint uses Overall. The title is Overall Agentic Readiness over time. The note under it is Workspace average across repositories. Repos without a scan yet count as 0%.

Readiness series

Next to Weekly / Monthly, Readiness series lists Overall first (the default), then the nine catalog names: Agentic Readiness, Token Efficiency, Repository Health, Codebase Quality, Tests & QA, Observability, API Contracts, Continuous Integration, and Continuous Delivery.

Changing the list updates the same chart in place. It does not open a dimension-detail page and does not change the URL. The choice lasts while you stay on the overview, same as Weekly/Monthly. It is not saved across visits.

A selected dimension changes the title to {catalog name} over time and the note to Workspace average of {catalog name} across repositories. Repos without a scan yet count as 0%. Example: pick Tests & QA and the title becomes Tests & QA over time. Pick Agentic Readiness and it becomes Agentic Readiness over time. Switch back to Overall to restore the first-paint title.

The line is the workspace average for the series you picked. Repos that have not been scanned, or that have no stored value for that dimension, count as 0%. Hover Overall to see the date plus the full dimension breakdown. Hover a single dimension to see the date and that one percent, labeled with the catalog name.

Weekly and Monthly still group the line you are looking at. Default is Weekly. If there are not enough buckets, switch granularity or run more scans.

Opening a dimension from the sidebar (or a URL like /agentic-foundation?dim=tests-qa) is a different page. That detail view has no workspace series selector.

Repositories table

The comparison table on Agentic Foundation is the list under Repositories. Dimension scores stay collapsed until you use Columns. The score header is Overall.

Default columns

These five columns always show, in this order. You cannot hide them. Scan status is not its own column. The Scan control sits on the Repository cell. See Analysis status.

Column	What it shows
Repository	Repo name. Language, platform, ineligible, size-limit, and warning chips stay on a second line here (or in the name tooltip). They do not move into the score columns.
Overall	The 0–100 readiness score and band after a real scan. Not scanned when the row has never been scanned.
Open issues	Open readiness issues on that repo.
Blocking rule	The one open gap or partial criterion on the latest successful scan that is not already fixed. Empty when the repository is not blocked. Sorting this column uses the full repository list, not only the current page.
Worst dimension	The lowest scored pillar on that row, as the catalog name plus the percent (example: Test Coverage 41%). Pillars marked N/A do not count. If two pillars tie, the table uses catalog order. Never scanned, or no scored pillar: —, not 0%.

Filter chips for language, platform, ineligible, and size stay in the strip above the table. They do not add columns.

Columns

Columns (toolbar control, “Choose dimension columns”) is a checkbox list of the nine dimension names. All nine start off. Extra columns you turn on append after Worst dimension, in catalog order.

Reset to default turns the extras off and returns the table to the five columns above. Extra columns stay for this browser tab and this workspace. A new tab starts with defaults again.

With only the five defaults, the table is a normal five-column layout. When any dimension column is on, the table can scroll sideways. On a medium-or-wider screen the repo name stays pinned on the left. On a phone-width screen it is not pinned, so the name does not cover the score cells.

Sort

The table lands with Overall lowest scored first. Click Overall again to flip to highest first.

Never-scanned rows sit after every scanned row in both Overall directions. They still show Not scanned / —, not 0%.

Open issues stays sortable on the page you are looking at. Worst dimension can use the same current-page sort. Blocking rule sorts the full list by severity. Repositories with no blocking rule stay last.

Empty states

Situation	What you see
Workspace has no repositories	No repositories analyzed yet (no table)
Search or chips match nothing	No repositories match your filters

The overall score

Each repository has one overall readiness score from 0 to 100. It rolls up the nine dimension scores with extra weight on pillars that matter most for day-to-day agent work (especially Agentic Readiness and Test Coverage). It is not a simple average of the nine percentages. It is also not the engineering delivery score on Insights → SDLC Insights.

Band	Range	What it usually means
Strong foundation	85–100	Strong docs, automation, and agent context.
Partial foundation	60–84	Solid base; a handful of targeted fixes left.
Material gaps	40–59	Meaningful gaps; check the Resolution Hub.
Not ready	0–39	Structural friction; prioritize top items.

Example: payments-api might show 58 (Material gaps) while handbook shows 82 (Partial foundation). The API repo has fixable gaps; the docs repo simply has a different shape of criteria (many delivery rows are N/A).

Dimension scores

Below the overall score in the detail view, you see all nine dimensions. Each row includes:

Field	What it tells you
Score %	How well this repo meets that pillar’s criteria.
Confidence	How much evidence the scan found. Low confidence means thin or missing signals, not necessarily a bad codebase.
Status badge	Same band colors as the overall score (Strong foundation / Partial foundation / Material gaps / Not ready).

Example: On web-checkout, Agentic Readiness might be 45% with low confidence because there is no .cursorignore and no agent instructions. Test Coverage might be 72% with high confidence because Jest files and a vitest.config.ts are easy to detect.

Reading the detail view

Click a repository to open three main areas.

Open in…

On the repository detail header (and on Code Insights repo detail), Open in… opens a menu:

Option	What happens
Cursor	Starts cloning the repository in the Cursor desktop app (Cursor must be installed).
Cursor Web	Asks you to confirm, then opens Cursor agents on cursor.com. Cursor Web uses the Cursor account and organization you are signed into there, which may differ from this workspace.
Default IDE	Starts cloning with the IDE registered for vscode:// links on your computer.
Claude Code	Copies the repository clone URL so you can open the project in Claude Code.

The control can remember your last choice on this browser (Remember my IDE). Use this when you want to fix gaps locally (for example after exporting remediations) instead of only working in the browser.

Dimension breakdown

Each dimension expands to show individual criteria with a status:

Status	Meaning	Quick example
Met	Fully satisfied.	AGENTS.md lists npm test and npm run build.
Partial	Present but incomplete.	README says “run tests” without an exact command.
Gap	Missing expected signal.	No .github/workflows/ in a service repo.
N/A	Doesn’t apply here. Excluded from the dimension score.	No Dockerfile expected in handbook.
Blocked	Couldn’t verify at scan time. Usually counts as not met.	A GitHub API call failed while reading branch protection.

Example: In payments-api → Agentic Readiness, you might see Gap on “Agent instructions” and Met on “Scriptable build & test” because the README has commands but there is no AGENTS.md.

Blocked usually counts as not met. When split infrastructure is on and a selected platform repo cannot be read, the checks for that role are left out of the dimension percent instead. Failing GitHub Actions runs do not move the CI setup rows. See Linked platform repositories and Four CI questions.

In the Dimension Breakdown header you can also select Export remediations to download outstanding gaps (open and PR-pending) as Markdown (.md) or PDF (.pdf), grouped by dimension. Each item includes rubric metadata, agent-oriented fix instructions, and remediation hints. Export is complementary to in-platform Fix all and the AI Resolution Hub. It does not replace them.

How to use the exported remediation file

Use the download when you prefer to fix locally in your IDE (or with an AI coding assistant) instead of queueing automated fixes in the platform.

1. 1
   Download

   On the repo detail view, open Dimension Breakdown and click Export remediations. Choose Markdown (.md) or PDF (.pdf). Files are named like readiness-remediations-<repo>-<date>.md or .pdf.

2. 2
   Open the repo in your IDE

   Clone or open the same repository the scan scored (default branch unless you know you are fixing elsewhere).

3. 3
   Attach or paste the Markdown

   Drop the file into your agent chat, attach it as context, or paste the sections you care about. Markdown is best for coding agents; PDF is the same content for sharing or printing. Each issue includes rubric metadata, agent instructions, and remediation hints.

   When the export is for every dimension, it can also include CI, CD, and API contracts sections for repos you selected under Linked platform repositories. A Continuous Integration export includes only CI. A Continuous Delivery export includes only CD. An API Contracts export includes only API contracts. Any other single dimension omits the block. Those sections are context, not scored gaps. The file asks the agent to add them to AGENTS.md when they are missing, and not to copy platform files into the service repo.

4. 4
   Apply what the file can change in code

   Most remediations are file or CI changes (docs, ignore files, workflows, specs). Those belong in a normal pull request you open from your IDE.

5. 5
   Handle settings outside the PR

   Some Continuous Integration governance items (G1–G5) need GitHub org/repo settings, Sonar linkage, or similar. A PR cannot fix those. Follow the in-app instructions from Fix it, or see Automated fixes and GitHub write access and the Contracts & governance patterns.

6. 6
   Merge and scan again

   Merge code changes to the default branch, finish any settings steps, then click Scan on the repository list or Run analysis on the repository page. Scores and hub rows update only after a completed scan. See Closing the loop.

Issues list

A flat list of every Gap, Partial, and Blocked item across all dimensions, sorted High → Medium → Low severity. Each line has a short description and a remediation hint. This is the same content the AI Resolution Hub pulls from, scoped to one repo.

Score history

On a repository detail page, after two or more completed scans, Score over time shows how that repo’s overall score moved. That chart is separate from the workspace progression chart on the Agentic Foundation overview (see Workspace score over time). Use the repo chart to confirm that a merged PR actually improved readiness.

Example: payments-api goes from 54 → 71 after adding AGENTS.md and a CI test job. The chart should show a step up on the scan date following that merge.

Example: reading one repo end to end

Walk through payments-api (Node/TypeScript API) as a new engineering lead would:

1. 1
   List view

   Overall 58 (Material gaps), last scanned yesterday. Worst dimension shows the weak pillar on that row (for example Agentic Readiness 41%). Use Columns if you want the other eight scores on this list.

2. 2
   Detail header

   Nine dimension tiles: Agentic Readiness 41% (low confidence), Test Coverage 68%, Delivery Readiness 75%, and so on.

3. 3
   Drill into Agentic Readiness

   Gap: no AGENTS.md. Partial: .cursor/rules/ exists but descriptions are empty. Met: README has npm test and npm run build.

4. 4
   Issues list

   Top item: “Add agent instructions” (High). Second: “CI does not run tests” (High). Jump to AI Resolution Hub to see the same items ranked against other repos.

When scores update

Scores refresh on every completed scan. Pushing a fix to GitHub does not change numbers until the next run finishes.

Trigger	What happens
Initial onboarding	First scan queues automatically for each visible repo.
Scheduled scan	About every 24 hours per connected repo.
Manual scan	Click Scan on the repository row, or Run analysis on the repository page.
Failed scan	Previous score stays; the row shows failed status until you retry.

Typical wait: 5–15 minutes after you trigger a scan. Large monorepos can take longer.

Analysis status

The repository list does not poll for a new status on its own. The label is whatever the page loaded last. The control sits on the Repository cell, not in its own column. Click Pending or Analyzing… to check whether the run has moved.

Label	Where	Meaning
Scan	Repository list	Scores are from the latest completed scan. Click it to queue another scan.
Pending	List and repository page	A scan is queued and has not started.
Analyzing…	List and repository page	A scan is running. Scores update after you check the status or reload the page.
Run analysis	Repository page	Starts the same scan as Scan on the list.

Queueing… shows on the list while the click that starts a scan is in flight. Checking… shows on the repository page while a status check is in flight.

Common questions

Why did only one dimension move?
Fixes are pillar-specific. Adding openapi.yaml lifts API & Contract Design; it won’t move Test Coverage unless you also wire tests.

Why did the overall score drop?
New Gap or Blocked rows appeared: a file was removed, GitHub governance checks failed, or the rubric now detects something that was invisible before. Check the issues list for what changed.

Why is overall “Partial foundation” but agents still struggle?
A single Gap in Agentic Readiness (no AGENTS.md, no ignore file) can hurt agents more than the headline number suggests. Read dimension scores, not just the total.

Next

• The 9 dimensions
• AI Resolution Hub
• Agentic Foundation overview