BC-Bench is a benchmark for measuring coding agents on real-world Microsoft Business Central development tasks written in AL. It helps compare models and agent configurations using reproducible evaluation data.
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.
npx skills add microsoft/BC-Bench --skill extensibility-request-implementgit clone --depth 1 https://github.com/microsoft/BC-BenchWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/microsoft/bc-bench/extensibility-request-implement)<a href="https://agentmods.dev/skills/microsoft/bc-bench/extensibility-request-implement"><img src="https://agentmods.dev/badge/skills/microsoft/bc-bench/extensibility-request-implement/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/microsoft/bc-bench/extensibility-request-implement"><img src="https://agentmods.dev/badge/skills/microsoft/bc-bench/extensibility-request-implement.svg" alt="Reviewed on agentmods" width="80" height="20"></a>- NVIDIA SkillSpector warn
SkillSpector: 1 finding, up to medium
These are SkillSpector’s own severities. On a checked sample its high-severity flags on skills were ~96% false positives — a documented command, a public API, a “never do X” rule — so we show them as a caution to read, not a verdict. Why →
- medium Excessive Agency · line 99 Skill enables autonomous high-impact decisions without human-in-the-loop verification. Critical operations (destructive commands, financial transactions, data deletion) should require explicit user confirmation.Fix: Add human-in-the-loop confirmation for destructive, irreversible, or high-impact operations. Never auto-execute commands that modify files, send data, or alter system state.
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.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00175 | $0.02649 |
| Opus 5.5 | $0.00070 | $0.01060 |
| Sonnet 5.5 | $0.00035 | $0.00530 |
| Haiku 4.5 | $0.00017 | $0.00265 |
Grade A, and why
extensibility-request-implement 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 1mo 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.
How it starts
The opening of the file, as written. The whole thing — 204 lines — stays where its author put it; the contents beside it link to each section on GitHub.
AI Extension Fix Skill
This skill implements one Business Central extensibility request entirely offline, inside the current repository. It reads the request text, adds the requested extension point (most often an integration event) to the standard code following the repository guidelines, and applies the change W1-first with propagation to same-named country/region layer files. It does not build, test, commit, or open a pull request. The change is left as unstaged edits in the working tree; the author (or the evaluation harness) reviews the resulting diff.
Critical Rules
- Text input only: The extensibility request is provided as plain text in the current conversation / prompt. Never fetch anything from an issue tracker or any external system, and never depend on labels, tickets, or online metadata.
- Single request: This skill implements one request per run. If more than one distinct request is provided, implement the first and warn.
- No writeback, no PR: Never commit, never push, never open or edit a pull request, and never run
any
ghcommand. The only deliverable is the set of local.aledits. - No build, no tests: Never build, publish, or run the app, and never create tests. Produce a change that is aligned with the guidelines and, to the best of static reasoning, compilable. Correctness is verified by the human reviewer on the resulting diff.
- Surgical changes: Make the smallest diff that satisfies the request. Follow the repository's code surgery guidelines (see Step 2). Do not reformat, rename, reorder, or refactor unrelated code.
- W1-first layering: When the file to change exists in the W1 base layer (
App/Layers/W1/...), make the change in W1 first, then apply the same change to every same-named counterpart file that exists in the other layers (Step 5). Only when the file exists solely in a specific country layer do you edit that layer directly and skip the propagation step. - AL files only: Only ever create or modify
.alsource files. Never edit, add, or delete any other file type (for example.json,.xml,.xlf/translations,.md,.txt, project or settings files, workflows, or build scripts). If a correct change would require touching a non-.alfile, STOP and report that instead of changing it. - No ticket-reference comments in code: Do not add
// issue #<n>-style comments or any external reference into the source.
What ships with it
1 file beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
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.
- 1mo ago First seen · 204 lines · 175 tokens per session scan A 4a9c21141a33
extensibility-request-implement is a skill published in the GitHub repository microsoft/BC-Bench (57 stars, last pushed today), licensed MIT. It adds 175 tokens to every session and 2,649 once invoked, about $0.0007 per session on Opus 5.5. 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.
Other skills, from other repositories
skill-agent-toolkit
Build, configure, and integrate Business Central agents using the AI Development Toolkit and Agent SDK. Triggers on Agent SDK, Agent Metadata Provider, IAgentFactory, IAgentMetadata, IAgentTaskExecution, ConfigurationDialog, Agent Task Builder, Agent Session, Copilot Capability, agent instructions, or agent setup in…
skill-agent-task-patterns
Agent SDK task integration patterns for Business Central. Triggers on: Agent Task Builder, Agent Task Message Builder, Public API, AssignTask, ExternalId, agent session detection, BindSubscription, TryFunction agent task, multi-turn agent conversation, agent attachment, task lifecycle management, OnPrem-only methods…
skill-api
AL API development patterns for Business Central. Use when creating OData/REST API pages, HttpClient integrations, webhook implementations, or any external system integration via API.
skill-copilot
AL Copilot capability development for Business Central. Use when implementing PromptDialog pages, AI generation features, or integrating with the Copilot toolkit.
skill-pages
AL page design patterns for Business Central. Use when creating Card, List, ListPart, Document pages, implementing page extensions, or designing UX layouts.
skill-testing
AL test development patterns for Business Central. Use when creating test codeunits, writing Given/When/Then test procedures, using Library Assert, configuring test projects, or implementing TDD workflows.