StefanMaron

Run mutation testing against an AL project using al-mutate (MSDyn365BC.AL.Mutate). Identifies test gaps by applying AST-based mutations and checking whether the test suite catches them via al-runner. No BC instance required. Produces mutations.json and report.md.

Download AL symbol packages (.app files) for the current project's dependencies from Microsoft NuGet feeds. Reads app.json to determine required packages.

Look up Business Central base application source code (tables, pages, codeunits, events) from MSDyn365BC.Sandbox.Code.History. Use this to verify object structures, find event publishers, or check field definitions.

Quick reference for AL build pipeline. Auto-load when compilation, deployment, or testing is mentioned. For running commands, use /compile, /publish, or /run-tests.

Compile the AL project using al-compile. Runs analyzers, reports errors, and writes diagnostics to the task folder.

Implement AL/BC solution using parallel development agents and 4-specialist review team. Spawns N developer agents for parallel modules, then 4 reviewer agents for comprehensive code review.

Generate comprehensive technical documentation for implemented AL features using a docs-writer specialist agent.

Lightweight bug fix workflow. 3-tier classification for fast iteration without approval gates.

Pipeline architecture, stage-by-stage flow, BC→mock type rewriting, and the key-file map for AlRunner. Use when modifying AlRunner/ source (Program.cs, RoslynRewriter.cs, Pipeline.cs, Runtime/**), debugging compilation/transpilation issues, adding new mock types, understanding why a BC type is or is not rewritten, or interpreting non-zero exit codes (1/2/3).

How to write, place, and run AL test suites in this repo — proving-test rules, bucket/category layout, AL object-ID uniqueness, run-all-buckets command, build commands, stubs invocation. Use when adding a test suite under tests/, picking object IDs for a new codeunit/table/page, running the test matrix locally, or evaluating whether an existing test "proves" anything.

Multi-agent workflow contract for this repo — orchestrator vs implementation agents, GitHub label state machine (status, agent), PR lifecycle, and the role of `al-runner --guide`. Use when acting as orchestrator/impl-agent without the dedicated sub-agent, when triaging the issue/PR queue manually, or when updating runner capabilities (must update PrintGuide() in lockstep).