fix: dashboard and CLI agree on what is deployed

[Robobc]

What

Usability-review finding 9: the dashboard contradicted reality three ways.

Arithmetic that could go negative. monitor.py classified by substring, so ROLLBACK_COMPLETE counted as both deployed ("COMPLETE") and failed ("ROLLBACK"), and not_deployed was computed by subtraction. Five healthy stacks plus one rolled back reported deployed=6, failed=1, not_deployed=-1. The browser classified the same status differently, so the terminal and the UI disagreed with each other.

A fake denominator. The UI used Math.max(stackCount, 10), so a healthy greenfield deployment read 6/10 and looked permanently incomplete.

Three sets of module labels. monitor.py said 6a/S2, index.html said Module 1, deploy.sh's MODULE_MAP says E.

Fixes

• Classification lives in monitor.classify() and only there. Each stack carries a state field and the browser renders it; the browser's own classifier remains as a fallback for older status.json files, with identical rules.
• Scope comes from the deployment contract — expected_stacks() decides what this configuration promises, federation-aware. Anything else is not-applicable, shown dimmed as "Not In Scope", never counted as missing or failed. The denominator is the contract's count.
• Summary counts are tallied from the emitted states rather than subtracted, so a negative count is now unrepresentable.
• monitor.py owns the module/team labels and the browser prefers them.
• Cleared monitor.py's pre-existing lint debt while touching it (CI lints whole changed files): sorted imports, no bare try/except/pass, no datetime.utcnow(), a logger instead of silent swallows.

Verified live

Against the running test environment:

total_stacks: 6   deployed: 6   in_progress: 0   failed: 0
not_deployed: 0   not_applicable: 4   (in-scope states sum to 6/6, no negatives)

That same deployment previously displayed 6/10 with four stacks looking absent.

Tests

tests/test_dashboard.py pins all three defects: rollback is failed and only failed, states are mutually exclusive, the denominator floor is gone, the UI prefers the emitted state, and module labels are checked against MODULE_MAP parsed out of deploy.sh — a real drift guard that fails if either side is renamed.

149 tests green, ruff clean.

[github-actions]

Commit: 1e55f84 | Updated: 2026-08-25 15:13:31 UTC

Security Scan Results

Scanner	S	C	H	M	L	I	Time	Action	Result	Thresh
bandit	0	0	0	0	16	0	853ms	0	PASSED	MED (g)
cdk-nag	0	0	0	0	0	0	8.2s	0	PASSED	MED (g)
cfn-nag	0	0	0	0	0	0	7ms	0	PASSED	MED (g)
checkov	0	0	0	0	0	0	4.9s	0	PASSED	MED (g)
detect-secrets	0	0	0	0	0	0	318ms	0	PASSED	MED (g)
grype	0	0	0	0	0	0	1m 2s	0	PASSED	MED (g)
npm-audit	0	0	0	0	0	0	187ms	0	PASSED	MED (g)
opengrep	0	0	0	0	0	0	18.2s	0	PASSED	MED (g)
semgrep	0	0	0	0	0	0	<1ms	0	MISSING	MED (g)
syft	0	0	0	0	0	0	2.6s	0	PASSED	MED (g)