al-agent-harness

Plugin al-agent-harness from marketplace al-agent-harness in waldo1001/DirectionsEmea.ALAgentHarness. Commands read from the repo's manifest — 13 days ago.

Skills (6)

• al-build-validation

  Build, diagnose and fix AL code until the build is green. Use when: you created or changed any .al file or app.json, before saying a task is done, before a review or a PR, and when the user asks "does it compile". Provides the build -> diagnostics -> fix -> rebuild loop and a fixed report. Code is n

• al-knowledge

  Find sourced Business Central / AL guidance instead of answering from memory. Use when: asked "how should I ..." about AL or BC, choosing a pattern (events, error handling, labels, permissions, performance, upgrade, testing), checking what an object, field, event or procedure is or does in this BC v

• al-object-id-assignment

  Reserve object, field and enum-value IDs with AL Object ID Ninja BEFORE writing the AL artifact, and release IDs that end up unused. Use when: creating any new AL object (table, tableextension, page, pageextension, codeunit, report, enum, enumextension, query, xmlport, permissionset, ...), adding a

• chat-export

  Export an agent session and turn it into a short diagnose report for whoever maintains the harness: which skills, agents and MCP servers were available versus actually used, which tools failed, where the agent looped. Use when: "export this chat", "why did it not use the skill", "send this session t

• company-rules

  Load and apply the company AL rules in rules/ as mandatory constraints, and report PASS / FAIL per rule. Use BEFORE writing or designing any AL code, when reviewing AL, and for /done line D7. Enumerates the folder every time: a new rule file applies without editing this skill.

• definition-of-done

  Check a change against DEFINITION-OF-DONE.md mechanically: every line gets a command or tool call as evidence, then one word, done or not done. Use when: /done, "is it done?", "can I open a PR?", before handing work to review, after the al-coder agent finishes. Never fixes: reports. D8 publishes to

Slash commands (4)

• /al-agent-harness:done

  Check the current change against DEFINITION-OF-DONE.md: evidence per line, then one word - DONE or NOT DONE. Never fixes; D8 publishes to the sandbox and runs the tests.

• /al-agent-harness:pr

  Write the pull request description from the spec and the /done report, and open the PR on request. Example wiring: GitHub (gh). Replace it with whatever your shop uses.

• /al-agent-harness:sdd

  Implement an approved spec test-first on the sandbox: the full al-sdd loop (spec -> test -> RED -> business logic -> GREEN -> mutation -> RED -> GREEN), then /done. Runs the al-sdd agent in this session, so it can publish, run tests and ask you.

• /al-agent-harness:spec

  Write the spec for an AL feature: reconnaissance, one round of design questions, specs/<feature>.md. Runs the al-spec agent in this session so its questions reach you.

Subagents (6)

• al-build-fixer

  Builds the AL apps, reads the diagnostics and fixes them until the build is green, in its own context, and returns only the short build report. Use when: 'does it build?', 'fix the build', after changing .al files or app.json, before /done. Does not run tests and does not change behavior to get gree

• al-coder

  Implements an approved spec in AL: reconnaissance, reserves every ID with Ninja, writes the objects and the tests, runs the build loop until green, sets the spec to implemented and hands off to /done. Use when: a spec in specs/ is approved and needs building, 'implement the spec', 'implement specs/<

• al-lookup

  Answers a simple Business Central / AL knowledge question on a cheaper model, in its own context, and returns only the answer with its sources. Use when: 'which event fires before ...', 'what is the pattern for ...', 'does property X exist in our BC version', a quick best-practice check. Not for des

• al-reviewer

  Reviews AL changes the way a senior developer reviews a PR: read-only, findings first, every finding tied to a rule or marked opinion, severity blocking / should-fix / nit, one-word recommendation. Use when: 'review my changes', before opening a PR, after /done is green, 'what would a reviewer say'.

• al-sdd

  Spec-driven, test-first AL implementation against a BC SaaS sandbox, the full loop: spec -> test -> RED (the test fails) -> business logic -> GREEN (the test passes) -> mutation (break the logic on purpose) -> RED (the test catches it) -> GREEN (revert, the test passes again). Use when: 'do this tes

• al-spec

  Writes the spec for an AL feature before any code exists: reads the task, does reconnaissance in the code and the symbols, asks the 3-4 questions that change the design in ONE round, and writes specs/<feature>.md with status draft. Use when: a new feature or change needs designing, 'write a spec', '

MCP servers (2)

• al-object-id-ninja

  runs: npx -y @vjeko.com/al-object-id-ninja-mcp

• al

  runs: al launchmcpserver --transport stdio

Hooks (2)

• SessionStart

  node "${CLAUDE_PLUGIN_ROOT}/hooks/session-start.mjs"

• PreToolUse

  node "${CLAUDE_PLUGIN_ROOT}/hooks/no-literal-ui-text.mjs"

Similar plugins