bcquality/docs/standalone-runner.md

Build a lightweight standalone review runner

Documentation | Architecture

BCQuality provides review knowledge, routing, execution instructions, and structured output contracts. It intentionally does not choose models, schedule agents, retry failures, or collect usage telemetry. A standalone runner can add those host-specific capabilities without copying Business Central rules out of BCQuality.

Use the built-in standalone plugin when the host's default execution is sufficient. Build a runner when you need explicit control over cost, latency, concurrency, or integration with another review surface.

Start with the minimal integration example to connect your agent to the content before adding runner-specific behavior.

Keep BCQuality current

For plugin installation, use the quick start. For version identifiers, forks, and reproducible snapshots, see updates and versions.

A runner that reads BCQuality from a checkout should pin a commit or release and upgrade it deliberately. Do not copy knowledge files or action-skill prose into the runner; doing so creates a second, drifting quality policy.

Review a complete app folder

For a committed app, generated fixture, or source tree that has no meaningful diff, supply the app's root directory as folder-path. The review scope is every relevant file below that directory, including app.json and AL source. The folder does not need to be a Git repository.

The app-review example uses this input through the standalone adapter. Because a folder is a current-state snapshot, the review must not invent a previous app version when evaluating comparison-only rules. Entry can return more than one top-level skill; preserve all reports, including separately dispatched Community reviews, rather than assuming the Microsoft coordinator is the only result.

Minimal runner flow

1. Give the agent the review input and a task context containing the user's actual goal, available input types, and any known BC applicability dimensions.
2. Invoke skills/entry.md. Entry prepares the knowledge index and returns the action skills to run. Do not reproduce its routing logic.
3. Execute every dispatched action skill with the exact input subset in its dispatch record. Read skills/read.md and skills/do.md on demand.
4. When an action skill declares sub-skills, resolve its ordered leaf slots with tools/Resolve-SkillWorklist.ps1, passing the enabled layers and disabled skill paths from the task context. Execute every resolved leaf as a discrete invocation. Leaves are independent and may be scheduled serially or concurrently. Before dispatch, preserve the final ordered selection and legitimate configuration/input-compatibility exclusions as a private expected composition artifact; see composition acceptance.
5. Capture the exact Task return as the immutable raw audit payload and primary transport. Preserve it unchanged in private artifacts or host logs. Before the full DO acceptance gate, create a normalized candidate only for DO's bounded citation-ID and optional-range cases after strict JSON, full structural schema, and reference-integrity checks. Record normalizedIds and removedRanges separately in private telemetry, and accept the candidate only if the entire copy passes the unchanged strict gate. Use tools/Validate-FindingsReport.ps1, passing the exact source paths and fully retrieved article paths; pass -SkillKind super and -ExpectedCompositionPath for the final rolled-up report. The accepted report contains no undeclared telemetry fields.
6. Collect each accepted findings-report into sub-results in the declared sub-skills order, not completion order. Run the super-skill self-review only after all leaves have finished.
7. Apply the DO composition, failure, deduplication, reference-integrity, and outcome rules. Return strict JSON before rendering it for people or another system.

The runner must never inspect the diff to skip a review domain. A leaf decides its own task-level applicability and reports not-applicable or no-knowledge.

Composition acceptance

The findings-report validator requires PowerShell 7.5 or later to preserve literal JSON strings with ConvertFrom-Json -DateKind String.

A report can be internally consistent while omitting a selected review. Bind the final acceptance gate to the host's selection, not just the returned reports. The private JSON input to -ExpectedCompositionPath follows the DO consumer acceptance contract. For example, if style is selected and security was disabled:

{
  "superSkill": { "id": "al-code-review", "version": 1 },
  "subSkills": [{ "id": "al-style-review", "version": 1 }],
  "skipped": [{ "id": "al-security-review", "version": 1, "reason": "configuration" }],
  "acceptedResults": []
}

Build this artifact from Resolve-SkillWorklist.ps1 and the input-compatibility decision before dispatch. Resolver skipped entries have no version: enrich them from the declared skill's indexed version. Move an input-incompatible selected slot to skipped with its selected version and not-applicable reason; never use source content or later model output to make that decision. Keep paths, layers, and other resolver metadata if useful for private audit. Do not add the artifact to the findings-report or overwrite it to hide an unfinished invocation. Preserve it with the run's raw payloads.

Initialize acceptedResults before dispatch. After accepting a leaf, save its exact accepted copy (including any permitted normalization) in an immutable host-owned file outside worker/composer write access. Append only its capture to acceptedResults, for example:

{"id": "al-style-review", "version": 1, "reportPath": "accepted/style.json"}

reportPath may be absolute or relative to the composition artifact's directory. Capture a host-created failed validation report the same way. Never construct these captures from the composed sub-results or permit the composing model to supply or alter them. Keep the original selection and exclusions unchanged.

The gate rejects repeated or unexpected leaf IDs, wrong selected versions, reordered results, fabricated or missing exclusions, and uncaptured or altered leaf content. JSON property order is immaterial; array order, field presence, types, and values must match. Every captured leaf must appear in sub-results. If selected leaves remain unfinished, the report must be partial with a non-failed returned report, otherwise failed; name every unfinished leaf ID exactly in outcome-reason. Top-level from-sub-skill: "agent" findings are rejected while selected leaf results are missing. Do not fabricate leaf reports or use configuration skips for budget exhaustion. Wait for started invocations to finish; omit the self-review if the selected composition remains incomplete. Coverage still sums the non-failed leaf knowledge worklists, not the number of selected leaf slots.

Calls without the expected artifact remain supported for structural/semantic validation, but cannot certify that a composed review covered its selection. They also cannot bind nested leaves to the host's accepted outputs. They still reject duplicate leaf IDs and returned-and-skipped conflicts.

Runner-owned choices

Keep these settings and behaviors outside BCQuality:

• coordinator and leaf models;
• serial or concurrent scheduling and maximum concurrency;
• retries, timeouts, and rate-limit handling;
• token, cost, duration, and actual-concurrency telemetry;
• conversion of the findings report into Markdown, annotations, or PR comments.

Model selection and requested concurrency are deployment choices, not review rules. Evaluate them against representative applications before making them a default. Report actual usage and concurrency only when the host exposes native evidence; do not infer them from the requested profile.

Failure and output checklist

A compatible runner:

• invokes every worklisted leaf exactly once unless a documented retry replaces a failed attempt;
• resolves same-ID leaf implementations by custom > community > microsoft, preserves declared slot order, and falls back when a higher layer is disabled;
• keeps leaf contexts isolated and passes only the inputs they declare;
• preserves each raw Task return unchanged for audit and distinguishes it from any normalized accepted copy;
• removes only an optional range whose positive integer bounds contain the primary line but start before it, and only when the complete report has no other defect and the finding has no suggested-code;
• records normalization only in private runner telemetry and never adds fields to the findings-report;
• rejects reversed, invalid, or out-of-bounds ranges, range mismatches attached to suggested-code, and every repair outside DO's bounded exception;
• preserves every leaf report, including failed reports, in sub-results;
• excludes unreliable findings from failed leaves and returns partial when only part of the review is reliable;
• orders sub-results by the declared worklist and orders rendered findings deterministically;
• validates composition against the host-owned selection, and reports selected leaves left unfinished as incomplete rather than clean or configured away;
• calculates top-level severity counts from deduplicated top-level findings, not by summing leaf counts;
• preserves knowledge paths verbatim and verifies references before publishing;
• records the BCQuality commit or release used for the run.

A CI integration, custom agent, or small host-native plugin can implement this runner contract. These remain optional consumers: BCQuality's knowledge and skills stay independent of their orchestration choices.