Add microsoft-365 domain skill

[cipher982]

Summary

• Add Microsoft 365 domain skill covering SharePoint search, auth-cookie sync, and Loop viewer extraction.
• Document the raw .loop download trap and ServerRedirectedURL navigation path.

Validation

• Verified against a logged-in M365 tenant: SharePoint search returns rows, Loop viewer extracts decision text.

---

Summary by cubic

Adds Microsoft 365 domain skill docs for SharePoint Online and Loop to help agents authenticate, search, and read Loop docs without download traps. Adds a browser harness domain assertion to fail fast when the active tab is on the wrong domain.

• New Features
  • Auth cookie sync across Entra alias hosts with modern CDP fields to preserve.
  • SharePoint search via /_api/search/query from a same-origin tab; use ServerRedirectedURL for navigation.
  • Loop docs: avoid raw .loop links; use loop.cloud.microsoft viewer and simple text extraction.
  • Fast checks: auth-redirect probe and assert_page_domain to validate cookies and target site before actions.

^{Written for commit 19a6ffa. Summary will update on new commits.}

[cubic-dev-ai]

No issues found across 1 file

_{Re-trigger cubic}

[cubic-dev-ai]

1 issue found across 1 file (changes from recent commits).

Prompt for AI agents (unresolved issues)

Check if these issues are valid — if so, understand the root cause of each and fix them. If appropriate, use sub-agents to investigate and fix each issue separately.


<file name="src/browser_harness/helpers.py">

<violation number="1" location="src/browser_harness/helpers.py:190">
P1: `assert_page_domain` passes when both `expected_domain` and `host` are empty, so a falsy/unset expected value combined with a hostless page (e.g. `about:blank`) silently succeeds instead of failing closed.</violation>
</file>

<sub>Reply with feedback, questions, or to request a fix.

Fix all with cubic | Re-trigger cubic</sub>

[cubic-dev-ai]

P1: assert_page_domain passes when both expected_domain and host are empty, so a falsy/unset expected value combined with a hostless page (e.g. about:blank) silently succeeds instead of failing closed.

Prompt for AI agents

Check if this issue is valid — if so, understand the root cause and fix it. At src/browser_harness/helpers.py, line 190:

<comment>`assert_page_domain` passes when both `expected_domain` and `host` are empty, so a falsy/unset expected value combined with a hostless page (e.g. `about:blank`) silently succeeds instead of failing closed.</comment>

<file context>
@@ -175,6 +175,28 @@ def page_info():
+        raise RuntimeError(f"assert_page_domain: page has pending dialog: {info['dialog']}")
+    host = (urlparse(info.get("url") or "").hostname or "").lower()
+    expected = (expected_domain or "").removeprefix("*.").removeprefix(".").lower()
+    ok = host == expected or (allow_subdomains and host.endswith(f".{expected}"))
+    if not ok:
+        raise RuntimeError(
</file context>