mcpbeat

Diagnose Deployment

microsoft/diagnose-deployment

>- Surfaces PAC CLI upload errors and Dataverse async operation errors, pattern-matches against a known failure catalog, and optionally auto-fixes identified issues. Use when deployment errors", "fix deployment issues", "show upload logs", "why did my deploy fail", or "troubleshoot upload".

3k tokens
context cost
the whole folder, loaded on every use
1
files
instructions only
0
copies elsewhere
how many repositories repackaged it
601
stars on the repo
on the repository, not the skill itself

Install

one command, takes just this skill from the repository
npx skills add https://github.com/microsoft/power-platform-skills --skill diagnose-deployment

The instruction itself

15 sections, as written by the author

> Plugin check: Run node "${PLUGIN_ROOT}/scripts/check-version.js" — if it outputs a message, show it to the user before proceeding.

diagnose-deployment

Surfaces and pattern-matches deployment errors against a known failure catalog. For each identified error with an available auto-fix, asks explicit user permission before applying any changes. Never auto-applies fixes without confirmation.

Prerequisites

  • Project root with powerpages.config.json
  • PAC CLI installed (will report if missing)

Phases

Phase 1 — Verify Prerequisites and Locate Project

Create all tasks upfront at the start of this phase.

Tasks to create:

  • "Verify prerequisites and locate project"
  • "Collect deployment artifacts"
  • "Surface upload errors"
  • "Query solution import status"
  • "Diagnose and categorize findings"
  • "Offer auto-fixes"
  • "Present findings summary"

Steps:

  • Locate project root: search for powerpages.config.json in cwd and parent directories
  • Check PAC CLI: pac --version (report version or "not installed")
  • Check PAC CLI auth: pac env who (report environment URL or "not authenticated")
  • Check Azure CLI: az account show (report subscription or "not logged in")

Auth failures are non-blocking — report them as findings, continue collecting other artifacts.

Phase 1.5 — Ground in current ALM documentation

> Reference: ${PLUGIN_ROOT}/references/alm-docs-grounding.md

Cap this step at ~30 seconds. If MCP search / fetch errors out, log a one-line note and continue — this skill must remain runnable offline.

  • Run microsoft_docs_search with the query: Power Pages deployment errors solution import troubleshooting.
  • Fetch https://learn.microsoft.com/en-us/power-platform/alm/solution-concepts-alm (and at most one sister page on troubleshooting or known import errors) in parallel via microsoft_docs_fetch.
  • Extract a one-paragraph summary of what Microsoft Learn currently says about common deployment failures and their resolution. Compare against ${PLUGIN_ROOT}/references/deployment-error-catalog.md and flag any new error patterns not yet captured in the catalog.
  • Use the summary to inform pattern-matching in Phase 5. If a new pattern is documented on Learn that isn't in the catalog, surface it to the user as a candidate addition rather than silently extending the catalog.

Phase 2 — Collect Deployment Artifacts

Gather all available context:

  • Read powerpages.config.json — extract siteName, websiteRecordId, compiledPath
  • Check .powerpages-site/ folder exists
  • Glob for manifest files: .powerpages-site/*-manifest.yml — list all found, note their environment hostnames
  • Check if .solution-manifest.json exists (for solution-related diagnostics)
  • Check if docs/alm/last-import.json exists (for recent import failures)
  • Check build output: confirm {compiledPath}/ exists and is non-empty

Report: "Found project: {siteName}. Artifacts collected."

Phase 3 — Surface Upload Errors

Re-run pac pages upload-code-site in capture mode to get fresh error output:

pac pages upload-code-site --rootPath "." 2>&1

> Note: This intentionally triggers the upload to capture any errors. If the upload succeeds cleanly, that is also a valid diagnostic result ("no errors found").

Capture stdout+stderr as a single string. Pass to scripts/parse-deployment-errors.js:

echo "{escaped-output}" | node "${PLUGIN_ROOT}/scripts/parse-deployment-errors.js"

Parse the JSON findings array. If the upload succeeded with no errors, note this and skip to Phase 5 with an empty findings list.

Phase 4 — Query Solution Import Status

Only run if .solution-manifest.json exists.

  • Acquire Azure CLI token for environment URL
  • Check recent async operations (last 24 hours):
   GET {envUrl}/api/data/v9.2/asyncoperations?$filter=statecode eq 3 and statuscode eq 31 and createdon gt {yesterday}&$select=asyncoperationid,name,message,friendlymessage,statuscode,completedon&$orderby=completedon desc&$top=5
  • Check recent import jobs:
   GET {envUrl}/api/data/v9.2/importjobs?$select=solutionname,completedon,progress&$orderby=completedon desc&$top=3
  • If failed operations found, pass each message field through parse-deployment-errors.js

Skip gracefully if auth is not available (auth failure in Phase 1).

Phase 5 — Auto-Diagnose Known Issues

Consolidate all findings from Phases 3 and 4. For each finding, categorize:

  • Error: Blocks deployment
  • Warning: May cause issues
  • Info: Informational

Also add findings for missing artifacts discovered in Phase 2:

  • Missing websiteRecordId → Error (patternId: missing-website-record-id)
  • Empty build output → Error (patternId: empty-build)
  • Multiple environment manifests → Warning (may indicate environment confusion)

Present all findings in a table:

| # | Severity | Type | Issue | Auto-fix? |

|---|---|---|---|---|

| 1 | Error | upload | JavaScript uploads blocked | Yes |

| 2 | Warning | config | Multiple manifest files found | No |

Phase 6 — Offer Auto-Fixes

For each Error finding with autoFixAvailable: true, in order:

<!-- gate: diagnose-deployment:6.auto-fix | category=consent | cancel-leaves=nothing -->

> 🚦 Gate (consent · diagnose-deployment:6.auto-fix): Per-finding consent before applying any auto-fix. Loops once per Error finding with autoFixAvailable: true — each finding gets its own Yes / No / Skip-all AskUserQuestion. The pattern ID surfaces in the prompt. Never batch fixes — three findings = three separate consent prompts (unless the user picks "Skip all" on the first, which short-circuits the loop). The Yes from finding 1 does NOT cover finding 2; each fix has its own blast radius (different files, different settings, different reversibility).

  • Explain the issue and proposed fix
  • Ask explicit permission via AskUserQuestion:

> "Issue: {message}

> Proposed fix: {suggestedFix}

> Apply this fix? Yes / No / Skip all auto-fixes"

  • If approved, execute the fix:

stale-manifest: Delete *-manifest.yml file(s) in .powerpages-site/

   # Ask which manifest to delete if multiple found, then:
   rm ".powerpages-site/{manifestFile}"

blocked-js: Update blockedattachments setting

   # Get current setting first
   pac env list-settings | grep -i blocked
   # Remove .js from the blocked list (preserve other blocked types)
   pac env update-settings --name blockedattachments --value "{updated-value}"

missing-website-record-id: Retrieve and update record ID

   pac pages list
   # Parse output, find matching site by name, then update powerpages.config.json

auth-expired: Guide re-authentication

   pac auth create --environment "{envUrl}"
   az login

empty-build: Run build

   npm run build
  • After each fix, re-run the relevant check to verify it resolved the issue. Update finding status to "Fixed" or "Manual required".

> Key Constraint: Never apply any fix without explicit user permission. Each fix requires a separate confirmation.

Phase 7 — Present Findings Summary

Display a final summary table of all findings:

| # | Severity | Type | Issue | Status |

|---|---|---|---|---|

| 1 | Error | upload | JS uploads blocked | Fixed |

| 2 | Error | config | Missing websiteRecordId | Manual required |

| 3 | Warning | config | Multiple manifest files | Informational |

Status values:

  • Fixed: Auto-fix was applied and verified
  • Manual required: No auto-fix available — show manual steps
  • Skipped: User declined the fix
  • Informational: Warning/Info, no action needed

If all errors are resolved: suggest retrying deployment with /power-pages:deploy-site.

If manual steps remain: list them explicitly with commands or links.

Record Skill Usage

> Reference: ${PLUGIN_ROOT}/references/skill-tracking-reference.md

Follow the skill tracking instructions in the reference to record this skill's usage. Use --skillName "DiagnoseDeployment".

Key Decision Points (Wait for User)

  • Phase 6: Each individual auto-fix requires explicit confirmation before applying
  • Phase 6: If user says "Skip all auto-fixes", stop offering fixes and go to summary

Error Handling

  • If pac pages upload-code-site produces no output: report "No upload errors detected in current state"
  • If Azure CLI token unavailable: skip Phase 4, note in summary
  • If parse-deployment-errors.js returns no findings: report "No known error patterns detected" and show raw output for manual review

Progress Tracking Table

| Task subject | activeForm | Description |

|---|---|---|

| Verify prerequisites and locate project | Verifying prerequisites | Check PAC CLI, Azure CLI, locate project root and powerpages.config.json |

| Collect deployment artifacts | Collecting deployment artifacts | Read config, list manifests, check build output, check solution manifest |

| Surface upload errors | Surfacing upload errors | Re-run pac pages upload-code-site in capture mode, parse stdout/stderr |

| Query solution import status | Querying solution import status | Check recent failed async operations and import jobs in Dataverse |

| Diagnose and categorize findings | Diagnosing findings | Pattern-match all errors against deployment-error-catalog, assign severity |

| Offer auto-fixes | Applying auto-fixes | For each fixable error, ask permission and execute fix, verify result |

| Present findings summary | Presenting summary | Show all findings table with severity and fix status, list manual steps |

How to use it

Copy the folder

Take microsoft/diagnose-deployment from the repository into ~/.claude/skills for personal use, or into .claude/skills inside a project.

Check the name does not clash

The agent identifies a skill by the name field in its header. Two skills with the same name cannot sit side by side — one of them will be ignored.