/al-agent-instructions

description: Generate optimized natural language instructions for Business Central agents (Designer or SDK). Applies the Responsibilities-Guidelines-Instructions framework from skill-agent-instructions. Output stored in .resources/Instructions/InstructionsV1.txt for SDK agents. ALDC workflow (Copilot prompt al-agent.instructions); invoke explicitly.
disable-model-invocation: true

> **Copilot CLI adapter — generated; do not edit this distribution.** > Select a role using /agent or copilot --agent <id>. Role names in this contract > are routing destinations, never chat mentions or evidence of an invocation. > Delegate only through the native task tool, using the exact discovered agent ID > as agent_type (for example al-planning-subagent). Inspect list_agents and the > actual task schema first. Keep all canonical human gates. Pass bounded context > inline; wait for completion (read_agent for a background task) before consuming > the result. A delegated role returns questions to its caller for the human gate. > If task or the requested custom agent is unavailable in this context, return > the blocked handoff to the caller/user. Never impersonate a missing subagent, > silently substitute general-purpose, or launch another CLI process to fake it. > Tool aliases in frontmatter select capabilities; call the actual tools exposed > by the host: view, glob, grep/rg, edit/create/apply_patch, bash/powershell, > web_fetch, task. Use a capability only when granted to this role and available. > Load domain skills through the host's skill mechanism when exposed, otherwise > read the complete bundled SKILL.md. A read is not a native skill invocation. > Resolve PLUGIN_ROOT to this installed agent's plugin directory (not the project > or original checkout). It is a path placeholder here, not a promised global > shell variable. Read docs/copilot-cli-al-tooling.md there for LSP/MCP bindings; > it supersedes older availability examples in the shared terminal contract at > skills/skill-migrate/references/cli-al-tools.md, never role duties or human gates. > Project instructions and matching .github/instructions rules remain binding; > pass the relevant excerpts to delegated roles. Canonical artifacts remain in > .github/plans/. Role write scopes are behavioral, not filesystem sandboxes. > AL execution requires a verified terminal command/runner or an actually exposed > MCP capability. Reuse the configured AL LSP for Agents wrapper through native > CLI semantic tools when exposed to this role. Inspect the catalog; do not invent > tool aliases or broaden read-only grants to enable rename/write operations. > Editor debugger controls remain separate from semantic navigation. Official AL > MCP selectors use al/<tool>; only Developer/Implementer receive compile/build/ > restore. Discover the actual callable names and pass the correct App/Test path. > No official publish/authentication tool is granted by this adapter. Triage alone > receives the optional dedicated bc-profiling and bc-snapshot proxies; discover > their schemas and confirm an authorized target/window before capture. Other > roles consume evidence. Capture does not prove autonomous snapshot debugging. > Symbol MCP: before any tools/call, verify AL CLI prerequisites are already > provisioned; this provider may auto-install AL tools on first use. A read-only > role must return that prerequisite to the caller, never bootstrap software. > Load symbols via al_packages(action: load, path: absolute consumer path). > The MCP process may run from the plugin cache; never assume its cwd is the > consumer. Use the discovered schema and report missing .app packages explicitly. > Record missing capabilities and unexecuted checks explicitly; never simulate.

Workflow: Generate Agent Instructions

Applies the Responsibilities-Guidelines-Instructions framework from `skill-agent-instructions` to produce a runtime-ready instruction file. This prompt does not contain the framework — it executes it.

**Load skill first**: `skill-agent-instructions` (framework, keywords table, validation checklist, anti-patterns).

Output modes

• **Designer mode** → text ready to paste into the Agent Designer wizard
• **SDK mode** → text stored in `.resources/Instructions/InstructionsV1.txt`, loaded via `NavApp.GetResourceAsText()` returning `SecretText`, applied via `Agent.SetInstructions(UserSecurityId, InstructionsText)`

Workflow

1. Gather inputs

From the developer:

1. **Agent purpose** — one-sentence business goal 2. **Pages in scope** — pages included in the agent's profile 3. **Fields to read/write** — exact field names from those pages 4. **Actions to invoke** — exact action captions on those pages 5. **Decision criteria** — business rules for branching (if/then) 6. **Output format** — how should the agent report results? 7. **Safety gates** — which actions need user intervention?

2. Draft RESPONSIBILITY

One sentence with business domain + outcome. No paragraphs.

3. Draft GUIDELINES

3–7 cross-task rules. Required: at least one **ALWAYS**, one **DO NOT**, one safety gate for critical actions, one data-access boundary (read-only vs read-write).

4. Draft INSTRUCTIONS per task

For each task:

1. Start with navigation (how does the agent reach the data?) 2. Read and **MEMORIZE** context before decisions 3. Apply business logic with explicit if/then branches 4. Execute actions using proper keywords (see `skill-agent-instructions` keywords table) 5. Report results with structured **Reply** format 6. Include error handling at the end of the task

5. Validate

Run through the validation checklist from `skill-agent-instructions`. Key items:

• [ ] Page/field/action names match agent profile exactly
• [ ] **MEMORIZE** placed BEFORE value is needed in later steps
• [ ] **MEMORIZE** includes example format
• [ ] Critical actions gated by user intervention or review
• [ ] Written in English
• [ ] Environment-agnostic (no hardcoded company names, URLs,