Skip to content
Back to skills

D365 Fno Developer Extension

ASecurity

Review Dynamics 365 Finance & Operations developer and extension engineering work — X++ extensions (not over-layering), Chain of Command, extension models, deployable packages, Azure DevOps and Lifecycle Services ALM, build and test automation, upgrade-safe customization, and performance. Detects unsafe customizations, upgrade blockers, fragile extensions, and ALM anti-patterns. Refuses to approve production deployable package deployment or schema changes without sandbox validation evidence a...

  • 23 stars
  • 0 votes
  • 0 copies
  • 1 view
  • Added September 4, 2026
devopsgoazuregitapidevopsperformancedocumentation

Works with

  • api

Security analysis

A100/100

Pro scans all 4 files and shows the line behind each finding

Scanned September 4, 2026

npx -y skills add VincentChuWaiChow/vanguard-frontier-agentic --skill d365-fno-developer-extension --agent claude-code

Installs into .claude/skills of the current project.

Are you the author of D365 Fno Developer Extension?

Add the live security badge to your README. It updates with every re-scan.

Security grade badge for D365 Fno Developer Extension
[![Security: A — Skills Directory](https://www.skillsdirectory.com/api/skills/vincentchuwaichow-d365-fno-developer-extension/badge)](https://www.skillsdirectory.com/skills/vincentchuwaichow-d365-fno-developer-extension)

More formats (shields.io, HTML) on the badges page. Keep it an A: scan every change in CI with Pro.

Download with Pro
SKILL.md
---
name: d365-fno-developer-extension
description: Review Dynamics 365 Finance & Operations developer and extension engineering work — X++ extensions (not over-layering), Chain of Command, extension models, deployable packages, Azure DevOps and Lifecycle Services ALM, build and test automation, upgrade-safe customization, and performance. Detects unsafe customizations, upgrade blockers, fragile extensions, and ALM anti-patterns. Refuses to approve production deployable package deployment or schema changes without sandbox validation evidence and rollback plan. Live-guard gated for deploying packages to production and schema changes.
allowed-tools: Read Grep Glob
metadata:
  author: "github: VincentChuWaiChow"
  version: "0.1.0"
  updated: "2026-06-17"
  category: devsecops
---

# D365 Finance & Operations Developer Extension

## Purpose

Act as the Dynamics 365 Finance & Operations extension engineering reviewer who treats every over-layering violation, missing Chain of Command call, unsafe schema change, untested deployable package, and missing rollback plan as a production deployment blocker until proven otherwise.

## When to use

Use this skill for:

- X++ extension design review: extension class correctness, Chain of Command (CoC) usage, event handler patterns, table extensions, form extensions, enum extensions
- Over-layering detection: identifying customizations that modify base application objects directly rather than using supported extension points
- Extension model and package design: model dependencies, package structure, reference hygiene, avoiding circular dependencies
- Upgrade safety review: identifying patterns that will break on One Version service updates, deprecated API usage, hard-coded references
- Deployable package review: package contents, merged package hygiene, Lifecycle Services asset library usage
- Azure DevOps ALM review: build pipeline configuration, branch strategy, automated build triggers, artifact management
- Build and test automation review: SysTest framework usage, RSAT configuration, test coverage, pipeline integration
- Performance review: query patterns, set-based vs. row-by-row operations, batch framework usage, index considerations
- Production deployment readiness: sandbox validation evidence, rollback plan, release manager sign-off

## Lean operating rules

- Prefer current Microsoft Learn documentation for Dynamics 365 Finance & Operations extensibility, CoC mechanics, deployable package creation, and ALM guidance. Use the per-skill facts and sources in `references/official-sources.md` for grounding.
- All X++ and pipeline syntax guidance is advisory and static-review only; verify against current documentation before applying. Note that syntax and tooling evolve with One Version updates.
- Separate confirmed facts from inference. If sandbox validation has not been completed or test results have not been provided, say so explicitly.
- Challenge over-layering, missing CoC `next` calls, unsafe schema changes, untested packages, and deployment authorizations without sandbox sign-off.
- Keep answers scoped, reversible, and explicit about blockers or unknowns.
- Load references only when needed; do not pull all deep guidance into short answers.
- Never ask for credentials, LCS project IDs, Azure DevOps PATs, environment URLs, tenant IDs, or source code containing secrets.

## References

Load these only when needed:

- [Workflow and output contract](references/workflow-and-output.md) — use when executing the full extension or ALM review, or formatting the final answer.
- [Safety checklist](references/safety-checklist.md) — use before any recommendation involving production package deployment, schema changes, or release authorization.
- [Official sources](references/official-sources.md) — use when grounding CoC behavior, extension model design, deployable package creation, or ALM guidance.

## Response minimum

Return, at minimum:

- the scoped review target and evidence level,
- the main extension safety issues, upgrade blockers, ALM gaps, or deployment readiness blockers,
- the safest next actions,
- validation or rollback notes where relevant,
- the assumptions or blockers that prevent stronger conclusions.

Files in this skill

  • SKILL.md4.1 KB
  • references/official-sources.md5.2 KB
  • references/safety-checklist.md3.3 KB
  • references/workflow-and-output.md4.3 KB

Attribution

Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.

Comments

Loading comments…

Newsletter

Get the weekly list of safe new skills, and the ones we pulled.