debug

A troubleshooting guide for a Claude add-in deployment in Microsoft 365. It routes common problems such as old configuration, connection failures, missing add-ins, sign-in loops, and local manifest testing.

In plain words
What is it for?
Use it to investigate why an add-in is not visible, cannot connect, shows stale settings, or fails during sign-in, including opening browser developer tools when needed.
Why use it?
It turns a vague deployment problem into a specific diagnostic path and uses copied error details when available.

Command

Install

Getting it into your agent

One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.

agentmods
npx agentmods add commands/anthropics/financial-services/debug
Clone the repo
git clone --depth 1 https://github.com/anthropics/financial-services
Per session 16 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 3,921 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. Scan, not verified.
Origin original No closer match found in the catalogue.
Token cost

What it costs to keep this loaded

Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.

ModelPer sessionOnce invoked
Fable 5 $0.00016 $0.03921
Opus 5 $0.00008 $0.01961
Sonnet 5 $0.00003 $0.00784
Haiku 4.5 $0.00002 $0.00392

Measured 2d ago against content hash 68128b967a22, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

debug scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 2d ago.

A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.

Nothing flagged

None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.

claude-for-msft-365-install/commands/debug.md · 359 lines

How it starts

The opening of the file, as written. The whole thing — 359 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Debug a Claude Office deployment

You are helping an enterprise admin diagnose why the deployed add-in isn't working right. Start by asking what's wrong, then route.

Triage

Ask the admin to describe the symptom. Route by answer:

Symptom Section
Updated the manifest but users still see old config Stale config after update
Add-in shows "Connection failed" Read the error paste
Add-in doesn't appear in Excel/PowerPoint at all Add-in not visible
Want to test/iterate a manifest locally before deploying Sideload a manifest for local debugging
Sign-in popup fails or loops Admin consent
Need to see the browser console Opening browser devtools

If they have an error paste from the add-in (the Copy error details button on the connect-failed screen), always start there. It carries everything.


Read the error paste

Paste structure:

Claude for Office connection failed (<Provider>)
Build: <sha>

<friendly message>

Request:
  <key>: <value actually sent>
  ...

Manifest params:
  <key>: <value the deployed manifest carries>
  ...

Raw error:
<SDK/HTTP error>

What to check:

  • Request: vs Manifest params: delta. Keys are the same snake_case names in both blocks, so diff directly. If they differ, the user typed override values into the form. If they match, the manifest values went through unchanged.
  • Manifest params: m key is the version tag (e.g. unified-1.0.0.11). If it's below what you last uploaded, the user is on a stale manifest. Go to Stale config.
  • Raw error: is the ground truth. Common patterns:
    • invalid_client (401, Google) → wrong google_client_secret for that google_client_id. Verify in GCP Console → Credentials.
    • Load failed (<host>) → network blocked at the WebView layer. Firewall needs to allow that host.
    • STS AssumeRoleWithWebIdentity failed → AWS IAM OIDC provider misconfigured or role trust policy wrong.
    • HTTP 401/403 (gateway) → bad token or gateway rejected the key.

Read the full file on GitHub · 359 lines

Changes

What this file has done since we first saw it

Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.

  1. 2d ago First seen · 359 lines · 16 tokens per session scan A 68128b967a22

Subscribe to this mod's changes

debug is a command published in the GitHub repository anthropics/financial-services (34,594 stars, last pushed 7d ago), licensed Apache-2.0. It adds 16 tokens to every session and 3,921 once invoked, about $0.0001 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.