feat: add monitor evidence inspect cli

scripts/src/hwlab-node-help.ts

@@ -83,6 +83,8 @@ export function hwlabNodeWebProbeHelp(): Record<string, unknown> {
       "bun scripts/cli.ts web-probe sentinel dashboard verify --node D601 --lane v03 --sentinel workbench-dsflash-go-tool-call-10x",
       "bun scripts/cli.ts web-probe sentinel dashboard screenshot --node D601 --lane v03 --sentinel workbench-auth-session-switch-2users",
       "bun scripts/cli.ts web-probe sentinel dashboard trigger --node JD01 --lane v03 --sentinel jd01-web-probe-sentinel",
+      "bun scripts/cli.ts web-probe sentinel inspect-url 'https://monitor.pikapython.com/?run=sentinel-run-xxxx'",
+      "bun scripts/cli.ts web-probe sentinel inspect-id webobs-xxxx --node JD01 --lane v03 --sentinel jd01-web-probe-sentinel",
       "bun scripts/cli.ts web-probe sentinel report --node D601 --lane v03 --sentinel workbench-dsflash-go-tool-call-10x --latest --view summary --raw",
       "bun scripts/cli.ts web-probe sentinel report --node D601 --lane v03 --sentinel workbench-dsflash-go-tool-call-10x --latest --view summary --full",
       "bun scripts/cli.ts web-probe sentinel maintenance stop --node D601 --lane v03 --sentinel workbench-dsflash-go-tool-call-10x --confirm --wait --release-id <id>",
@@ -105,6 +107,7 @@ export function hwlabNodeWebProbeHelp(): Record<string, unknown> {
       "collect views render bounded summaries from existing artifacts and do not create a second source of truth.",
       "performanceCapture records an explicit bounded Chrome CPU profile and drains LongTask/LoAF/event-loop-gap artifacts; performance-summary reads existing artifacts only.",
       "analyze is offline-only: it reads artifact JSONL and writes analysis/report.md plus analysis/report.json.",
+      "For monitor-web links or page evidence ids, start with `web-probe sentinel inspect-url` or `inspect-id`; default evidence must be controlled report/observe CLI output, not direct public API curl.",
       "`web-probe sentinel report --raw` returns bounded issue evidence JSON, including report/artifact SHA and root-cause signal summaries when available; use `--full` for the full indexed service payload.",
       "When multiple web-probe sentinels are declared, sentinel image/control-plane/validate/maintenance/dashboard/report require `--sentinel <id>`; plan/status without it show the registry drill-down.",
       "Issue evidence should cite observer id, stateDir, report SHA, screenshot SHA, command ids and concise summaries, not prompt/provider/secret payloads.",