[self-care:dashboard-review] Represent Central Agentic Ops in Packages and restore mobile Coverage access

Completed deterministic review against the trusted expected inventory, bounded Actions evidence, and the loopback-rendered dashboard artifact. I found 2 verified defects and 3 completed persona assessments: CFO **inefficient**, CSO **workable**, CTO **workable**. Workflow inventory matched the trusted inventory (`47/47`) and the dashboard honestly disclosed its partial/stale 5:55 AM run snapshot, so the priority work is to restore package-inventory correctness and mobile access to Coverage diagnostics.

**Action:** Assign to Copilot to represent the root `Central Agentic Ops` package in the Packages view while keeping the internal `dashboard` and `activity` packages excluded; accept when the Packages page shows 8 user-facing packages, still excludes those two internal packages, and preserves the current 47-workflow inventory.

### View grades
- `packages / packages-inventory`: overall `88.7`; clarity `1.00`, information signal `0.85`, cognitive economy `0.875`, visual balance `0.827`, legibility `0.75`. The heuristic stayed healthy, so the package omission was found by comparing the rendered table to the trusted expected inventory rather than by the composite score.
- `coverage / coverage-summary`: overall `89.1`; clarity `1.00`, information signal `0.85`, cognitive economy `1.00`, visual balance `0.693`, legibility `0.75`. Browser evidence at `390x844` found the summary table wider than its visible container and not given horizontal scrolling.
- `runs / runs-health-trend`: overall `90.0`; clarity `1.00`, information signal `0.725`, cognitive economy `1.00`, visual balance `0.96`, legibility `0.75`. This stayed usable overall, but the mobile trend labels support a follow-up usability suggestion rather than a verified defect.
- Dashboard build run: `34515402507`.
- Grader references carried forward: Shannon (1948) normalized entropy, Cleveland & McGill (1984) graphical perception, Mackinlay (1986) encoding effectiveness, Purchase (1997) graph readability, and Ngo et al. (2003) interface balance/density.

### Verified defects
1. **Packages inventory omits the root package.**
- Expected: the trusted `expected-inventory.json` for this run includes 8 user-facing manifests after excluding only the internal `dashboard/aw.yml` and `activity/aw.yml` packages: `Central Agentic Ops`, `AW Doctor`, `AW Optimization`, `Dependabot`, `Dev Practices`, `EU CRA`, `SelfCare`, and `UK AI Advisory`.
- Observed: the rendered **Packages** page shows `7 of 7` rows and omits `Central Agentic Ops`, while still listing the other 7 user-facing packages.
- Trusted comparison evidence: `expected-inventory.json` generated at `2026-09-10T20:52:46.142Z`; rendered Packages table extracted during this review.
2. **Coverage diagnostics clips key columns on a 390 px viewport.**
- Expected: the review contract requires desktop and `390`-pixel mobile views to remain usable without overlap or clipping.
- Observed: the rendered **Coverage diagnostics** table measured `642 px` wide inside a `360 px` visible container at `390x844`; the container reported `overflow-x: visible`, and the right-side `Coverage`, `State`, and `Reason` columns rendered beyond the viewport instead of inheriting the horizontally scrollable behavior used by the other tabular views.
- Trusted comparison evidence: deterministic browser measurements captured during this review; other mobile tables (`Packages`, `Repositories`, `Workflows`, `Dispatches`) exposed `overflow-x: auto` and remained horizontally scrollable.

Persona assessments

**CFO — skeptical**
- **Question:** How much AI Credit spend is concentrated in the top workflows, and is the operational value visible enough to justify that spend today?
- **Answer from rendered evidence:** Partially answered. The dashboard clearly surfaces attention and partial-data status, but it does not make cost/value evidence visible early enough for a fast CFO decision.
- **Path:** `#page-overview` → `#page-repository-detail?repository=githubnext%2Fgh-aw-cao` → `#page-coverage` → `#page-packages` → `#page-dispatches` → `#page-data-health` → `#page-overview`.
- **Evidence:** `8 items need your attention`; `6 failed runs`; `2 awaiting review`; partial-data notice; no visible AIC/value summary on Overview.
- **Efficiency:** `inefficient` — the persona could answer only part of the question without hopping across multiple views.

**CSO — optimistic**
- **Question:** Do the rendered dashboard views show any material security concerns or enough missing evidence that we should act now?
- **Answer from rendered evidence:** No active security findings are shown, but the dashboard also says security findings could not be fully evaluated in the available evidence window, so the absence of findings is confidence-limited rather than conclusive.
- **Path:** `#page-overview` → `#page-packages` → `#page-runs` → `#page-coverage` → `#page-data-health` → `#page-workflows`.
- **Evidence:** Overview shows `0 Security findings`; Overview also says security findings `could not be fully evaluated`; Data health and Workflows repeat the partial-data state.
- **Efficiency:** `workable` — enough evidence was available after retry, but partial/stale-state interpretation still required cross-view confirmation.

**CTO — hurried**
- **Question:** Where is the biggest DevOps infrastructure health risk, and is it severe enough to justify immediate engineering investment?
- **Answer from rendered evidence:** No acute infrastructure failure is surfaced, but the partial-data notice and incomplete run-health evidence prevent a strong investment decision from the dashboard alone.
- **Path:** `#page-overview` → `#page-repository-detail?repository=githubnext%2Fgh-aw-cao` → `#page-data-health` → `#page-packages`.
- **Evidence:** partial-data notice; repository detail AIC concentration led by `dependabot.md` at `3244 AIC (72.1%)`; package table shows `7 of 7` results and mostly zero-run/zero-AIC rows outside Dependabot and EU CRA.
- **Efficiency:** `workable` — the dashboard is quick to scan, but confidence limits are not prominent enough for a rushed technical decision-maker.

### Improvement suggestions
1. **Add the root `Central Agentic Ops` package to Packages with explicit bundle semantics.** This fixes the verified inventory omission and helps CFO/CTO readers understand the installed control-plane surface without surfacing the internal `dashboard` and `activity` packages. Encountered by CFO and CTO.
2. **Give Coverage diagnostics the same mobile horizontal-scroll treatment already used by the other tabular views.** This is the most localized way to restore access to the clipped right-side columns at `390 px`. Encountered during the deterministic mobile pass; aligns with CSO/CTO coverage review paths.
3. **Tighten confidence messaging when partial data suppresses security or cost conclusions.** Shorten repeated notices into one explicit confidence line near the top of Overview and security-adjacent pages, and surface a compact cost/value cue earlier for the CFO path. Encountered by CFO, CSO, and CTO.

Agent prompt
Represent the root `Central Agentic Ops` package in the dashboard Packages inventory as a user-facing bundle row while continuing to exclude the internal `dashboard` and `activity` packages. Reuse the existing expected-inventory and package-inventory data paths instead of inventing a new source. Preserve the current 47-workflow inventory and package-mode semantics. Update the Packages view so it shows 8 user-facing packages for this repository, with `Central Agentic Ops` clearly distinguished as the root bundle even though it has no direct workflow rows. Acceptance check: the Packages page renders 8 rows including `Central Agentic Ops`, the view still excludes `dashboard` and `activity`, and the Workflows page still reports 47 results.

Incomplete checks
None. All deterministic checks completed, and all three persona assessments completed with the CSO path succeeding on the allowed retry.

### Control Plane
- Correlation ID: `34528221498-543`
- Central repository: `githubnext/gh-aw-cao`
- Control plane run: https://github.com/githubnext/gh-aw-cao/actions/runs/34528221498

### References
- https://githubnext.github.io/gh-aw-cao/cao/
- https://github.com/githubnext/gh-aw-cao/actions/runs/34515402507
- https://githubnext.github.io/gh-aw-cao/cao/#page-coverage

> [!WARNING]
>
> Firewall blocked 2 domains
>
> The following domains were blocked by the firewall during workflow execution:
>
> - `mtalk.google.com`
> - `www.gstatic.com`
>
> To allow these domains, add them to the `network.allowed` list in your workflow frontmatter:
>
> ```yaml
> network:
> allowed:
> - defaults
> - "mtalk.google.com"
> - "www.gstatic.com"
> ```
>
> See [Network Configuration](https://github.github.com/gh-aw/reference/network/) for more information.
>
>

> Generated by [SelfCare / Dashboard](https://github.com/githubnext/gh-aw-cao/actions/runs/34528774719) · copilot · gpt54 · 201.4 AIC · ⌖ 7.95 AIC · ⊞ 24.8K · [◷](https://github.com/search?q=repo%3Agithubnext%2Fgh-aw-cao+is%3Aissue+%22gh-aw-workflow-call-id%3A+githubnext%2Fgh-aw-cao%2Fself-care-dashboard-review%22&type=issues)
> - [x] expires on Sep 24, 2026, 9:07 PM UTC

Start with the existing expected-inventory.json and package-inventory data paths, then inspect the Packages and Coverage diagnostics entry points. Reproduce the 7-row Packages result and the 642 px table at 390x844 before changing either view. Done means 8 user-facing packages including Central Agentic Ops, dashboard and activity excluded, 47 workflows preserved, and Coverage's right-side columns horizontally accessible on mobile.

Written by the indexing model from the issue text.