@apptoolkit/harness 0.0.0-stage → 1.0.0

package/dist/bun/native/payload/harness/workflows/W04.md

@@ -0,0 +1,13 @@
+# W04 Meeting preparation
+
+**Trigger**. Prepare my meeting or talking points.
+
+**Authority**. Use authorized calendar, notes and decision evidence. Scheduling, sharing and attendee notices are separate effects.
+
+1. Resolve the exact occurrence, audience, objective and available time. Retrieve the agenda, decision history, relevant changes and open questions.
+2. Distinguish previously decided items, today's decisions and information-only topics. Check attendance or roles only if the evidence supports them.
+3. Prepare concise talking points with evidence links, a proposed order and questions that need an owner. Run independent factual review for material technical claims.
+
+**Result and recovery**. Return an agenda-ready brief and unresolved items. If the meeting is ambiguous or a source is unavailable, identify the exact gap and continue with verified material.
+
+Apply the [operating policy](../POLICY.md) and [project guide](../PROJECT.md). Record reusable decisions and improvements at the appropriate checkpoint.

package/dist/bun/native/payload/harness/workflows/W05.md

@@ -0,0 +1,13 @@
+# W05 Transcript assimilation
+
+**Trigger**. Read this transcript and assimilate it.
+
+**Authority**. Read a supplied or authorized transcript or notes. Local knowledge changes remain scoped. External records and attributions require separate verification and authority.
+
+1. Resolve the exact occurrence, transcript intervals, speaker cues and completeness. Use a deterministic, format-aware reader for bounded excerpts and stable time locators.
+2. Separate accepted decisions and reversals, current facts, expectations, commitments and actions. Preserve uncertain speaker identity and missing intervals.
+3. Route decisions to history, facts to their indexed topic, commitments to the verified owner and actions to the active list. Verify each affected index and conflicting prior state.
+
+**Result and recovery**. Return decisions, actions, owners, locators and gaps. Preserve ambiguous items as pending with a concrete owner and resume condition.
+
+Apply the [operating policy](../POLICY.md) and [project guide](../PROJECT.md). Record reusable decisions and improvements at the appropriate checkpoint.

package/dist/bun/native/payload/harness/workflows/W06.md

@@ -0,0 +1,13 @@
+# W06 Decision history
+
+**Trigger**. What did we decide or record this in project history.
+
+**Authority**. Read scoped project decisions. Make only authorized local decision and index changes. Publishing or altering a shared record is separate.
+
+1. Resolve the authoritative decision record, status, dates, rationale, source locators and affected scope. Compare newer evidence and superseding records.
+2. Record an accepted decision with its evidence and owner. Keep proposals and unresolved questions distinct. Preserve prior versions and exact supersession links.
+3. Update canonical authority and relevant indexes together. Check links and recoverability from a fresh session. Use the independent audit for technical claims.
+
+**Result and recovery**. Return the current decision, rationale, effective scope and what it supersedes. If evidence conflicts, keep the conflict open and ask the proper decision owner.
+
+Apply the [operating policy](../POLICY.md) and [project guide](../PROJECT.md). Record reusable decisions and improvements at the appropriate checkpoint.

package/dist/bun/native/payload/harness/workflows/W07.md

@@ -0,0 +1,13 @@
+# W07 Weekly status report
+
+**Trigger**. Prepare the weekly report.
+
+**Authority**. Gather approved evidence and prepare a dated report. Publishing, mentions and announcements require their own grant.
+
+1. Fix the reporting window, time zone, audience, template and authoritative work sources. Collect tracked work, PRs, decisions and outcomes with coverage per source.
+2. Group completed, active and blocked work. Support each material claim with current evidence and distinguish an unpublished plan from a released result.
+3. Render the report and any requested announcement separately. Validate the exact template and links, independently audit technical claims and preserve the prior dated snapshot.
+
+**Result and recovery**. Return the dated report, optional draft and coverage gaps. Verify the index or authorized publication through readback when applicable.
+
+Apply the [operating policy](../POLICY.md) and [project guide](../PROJECT.md). Record reusable decisions and improvements at the appropriate checkpoint.

package/dist/bun/native/payload/harness/workflows/W08.md

@@ -0,0 +1,13 @@
+# W08 Proposal, release recap and metric
+
+**Trigger**. Prepare a proposal, release recap or metric.
+
+**Authority**. Read scoped artifacts and metrics and prepare the requested draft. Approval, publication and release activation are separate.
+
+1. Fix the audience, intended decision, artifact revision, date and reporting interval. For a metric, define the unit, query, numerator, denominator and exclusions.
+2. Collect representative measurements or release evidence. Separate proposals, completed outputs, activated changes and remaining work. Keep reported values tied to their windows and exact sources.
+3. Compare alternatives or baseline and result where appropriate. Audit material claims independently and validate the selected artifact before showing or publishing it.
+
+**Result and recovery**. Return the scoped draft and evidence. Report a missing denominator, collection gap or unverified serving revision as a pending claim.
+
+Apply the [operating policy](../POLICY.md) and [project guide](../PROJECT.md). Record reusable decisions and improvements at the appropriate checkpoint.

package/dist/bun/native/payload/harness/workflows/W09.md

@@ -0,0 +1,13 @@
+# W09 Reminder scheduling
+
+**Trigger**. Remind me to prepare the report or schedule a reminder.
+
+**Authority**. Read the scheduler capability and prepare an exact reminder. Creating, editing or removing a reminder requires a grant for that effect.
+
+1. Resolve the actual time zone, target date and time, recurrence, recipient, channel and end condition. Verify the scheduler supports them and distinguish a reminder from report generation.
+2. Preview the exact reminder text, timing and destination. Use the authorized scheduler operation only when approval covers the proposed effect.
+3. Read back the saved schedule where supported. Verify date, time zone, recurrence and delivery scope. Record verification of actual delivery separately.
+
+**Result and recovery**. Return the schedule identity and verification state. If the route cannot prove the required behavior, keep a usable draft and report the missing capability.
+
+Apply the [operating policy](../POLICY.md) and [project guide](../PROJECT.md). Record reusable decisions and improvements at the appropriate checkpoint.

package/dist/bun/native/payload/harness/workflows/W10.md

@@ -0,0 +1,13 @@
+# W10 Work log and timesheet
+
+**Trigger**. Recover my work log or fill this timesheet.
+
+**Authority**. Read only authorized user activity for the selected window. Creating a draft and submitting or correcting a timesheet have separate authorization.
+
+1. Confirm the local day, time zone, target format and approved sources. Retrieve bounded dated evidence for actual work and keep unrelated records outside the collection.
+2. Reconcile actions and elapsed intervals where supported. Distinguish attendance, work performed and estimated time. Preserve approved entries and find duplicates or conflicts.
+3. Prepare the exact entries and evidence, with gaps and uncertain attribution. Validate the template and preview any separately authorized correction or submission.
+
+**Result and recovery**. Return a scoped work log or draft. If proof is insufficient, mark the entry pending rather than inferring work from calendar or session metadata alone.
+
+Apply the [operating policy](../POLICY.md) and [project guide](../PROJECT.md). Record reusable decisions and improvements at the appropriate checkpoint.

package/dist/bun/native/payload/harness/workflows/W11.md

@@ -0,0 +1,13 @@
+# W11 Pull request review
+
+**Trigger**. Which PRs need review or review this PR.
+
+**Authority**. Start read-only. Resolve the authoritative forge or a supplied diff. Approving, commenting and assigning reviewers are separate effects.
+
+1. Resolve the repository, user identity, review queue criteria or exact change request and current base and head. For a live review, fetch discussions and checks first.
+2. Use bounded Git and diff tools to inspect changed behavior against accepted requirements. Check test evidence, negative cases, permission boundaries and existing discussions. Avoid duplicate findings.
+3. Report actionable findings with location, severity, expected and observed behavior, and supporting evidence. Tie conclusions to the inspected revision and identify skipped scope.
+
+**Result and recovery**. Return a review queue or revision-bound findings. If the head changes or essential discussions are inaccessible, qualify the result and use re-review when evidence returns.
+
+Apply the [operating policy](../POLICY.md) and [project guide](../PROJECT.md). Record reusable decisions and improvements at the appropriate checkpoint.

package/dist/bun/native/payload/harness/workflows/W12.md

@@ -0,0 +1,13 @@
+# W12 Pull request re-review
+
+**Trigger**. Re-review this change or check whether a thread is resolved.
+
+**Authority**. Read the prior findings, discussions and exact new revision. Resolving threads, approving and posting remain separate operations.
+
+1. Bind the previous finding and evidence to its revision and locator. Retrieve the current head, diff, discussion state and status checks.
+2. Reproduce the earlier concern or inspect the exact correction. Compare the contract and behavior and distinguish fixed, still present and unverifiable.
+3. Check for regressions in the changed scope. Prepare the exact next review action with the latest evidence and verify live readback after an authorized effect.
+
+**Result and recovery**. Return a status for each finding and the applicable head. If the head drifts, repeat the affected verification before relying on the result.
+
+Apply the [operating policy](../POLICY.md) and [project guide](../PROJECT.md). Record reusable decisions and improvements at the appropriate checkpoint.

package/dist/bun/native/payload/harness/workflows/W13.md

@@ -0,0 +1,13 @@
+# W13 Git writes and pull requests
+
+**Trigger**. Stage, commit, push or open a PR.
+
+**Authority**. Use Git for version control. Require separate permission for each exact write or an existing grant that covers it. A forge write is a separate effect.
+
+1. Inspect status, the intended files, base, head, branch and remote using Git read operations. Find the commit convention in the project guide or representative history.
+2. Verify the approved diff, applicable checks and literal commit text. Show the destination branch when it changes. For a PR, preview exact base, head, title, body, draft state and requested reviewers.
+3. Recheck the current head, authenticated actor, required permissions and destination before the effect. When automated review evidence gates delivery, verify both its producer and its revision. Execute only the covered effects, then inspect the staged content, commit identity, remote head or saved PR. Preserve user authorship and inspect conflicts before recovery.
+
+**Result and recovery**. Return exact resulting identities and open steps. If scope, target, hook results or remote state differs, stop that effect and report what needs correction.
+
+Apply the [operating policy](../POLICY.md) and [project guide](../PROJECT.md). Record reusable decisions and improvements at the appropriate checkpoint.

package/dist/bun/native/payload/harness/workflows/W14.md

@@ -0,0 +1,13 @@
+# W14 Outward message draft
+
+**Trigger**. Draft a message, announcement or reply.
+
+**Authority**. Draft in the user's voice. Saving to a shared system, posting and delivery require the exact applicable grant.
+
+1. Resolve audience, purpose, actor, approved sources and destination conventions. Check statements about the user against evidence or their input.
+2. Draft concise language with verified facts, links and mentions. Audit technical claims independently. Use the applicable deterministic title, body and metadata checks.
+3. Show the literal payload and destination for any external effect. If approved and supported, send or save through the native route and verify its actual identity and contents.
+
+**Result and recovery**. Return a ready draft or verified receipt. Preserve a local handoff when the service cannot represent the user's identity or the required payload.
+
+Apply the [operating policy](../POLICY.md) and [project guide](../PROJECT.md). Record reusable decisions and improvements at the appropriate checkpoint.

package/dist/bun/native/payload/harness/workflows/W15.md

@@ -0,0 +1,13 @@
+# W15 Ticket, bug and assignment preparation
+
+**Trigger**. Draft a ticket, bug, story, task or assignment.
+
+**Authority**. Prepare a local draft from approved evidence. Creating an item, changing state, linking records and assigning a person have separate effects.
+
+1. Resolve the project, canonical tracker, variant and applicable templates. Find existing issues, duplicates and dependencies. Verify the target's native metadata and allowed values.
+2. Prepare title, body, reproduction or acceptance criteria, links, owner proposal and required fields. Ground any proposed assignment in documented ownership and capacity. Validate exact contents.
+3. Preview each requested external effect. If authorized, perform only the corresponding write and read back fields, relationships and rendering. Report unexpected template or metadata changes.
+
+**Result and recovery**. Return the validated draft or verified item identity. Keep unclear routing or field constraints pending and preserve the usable evidence.
+
+Apply the [operating policy](../POLICY.md) and [project guide](../PROJECT.md). Record reusable decisions and improvements at the appropriate checkpoint.

package/dist/bun/native/payload/harness/workflows/W16.md

@@ -0,0 +1,13 @@
+# W16 Feature definition and technical decisions
+
+**Trigger**. Start a new feature, decide what to build, or prepare a specification, RFC or ADR.
+
+**Authority**. Read current authority and prepare the definition. Code changes require accepted, verifiable scope and authorization for implementation.
+
+1. Identify the user outcome, affected repository or shared workspace, current behavior, intended scope and decision owner. Read the authoritative ticket, specification and applicable local rules.
+2. If an existing authorized ticket or approved specification already settles the change, reuse it and map its requirements to checks. Otherwise define the feature with the user or designated decision owner. Resolve material ambiguities about behavior, exclusions, dependencies, effects and acceptance before coding.
+3. Record observable acceptance examples, testable behavior, scope exclusions, permissions, unresolved questions and the authority for each accepted choice. Validate the local artifact and separate proposal, acceptance and implementation authorization. Route accepted work to W17 or W18.
+
+**Result and recovery**. Return the accepted definition or the exact open questions. Continue authorized independent discovery while clarification is pending. Keep unresolved behavior out of implementation.
+
+Apply the [operating policy](../POLICY.md) and [project guide](../PROJECT.md). Record reusable decisions and improvements at the appropriate checkpoint.

package/dist/bun/native/payload/harness/workflows/W17.md

@@ -0,0 +1,13 @@
+# W17 Dependencies and work packets
+
+**Trigger**. Plan dependencies and split the work.
+
+**Authority**. Prepare a local plan or delegate within existing scope. Creating tracker items, changing ownership and dispatching work with new permissions are separate effects.
+
+1. Load the accepted W16 definition and split it by observable outcomes. Map dependencies, data and API boundaries, affected repository rules and needed permissions.
+2. Create a dependency graph, check for cycles and competing write paths, and name the owner, prerequisites, authorized paths, exact output and acceptance for each unit.
+3. Give each independent worker a self-contained packet with minimal evidence and verification. Include a coordinator for shared effects and confirm prerequisites before dependent work begins. If there is a budget, account for the coordinator and all workers in one ledger and preserve a verification reserve.
+
+**Result and recovery**. Return ordered units with ownership and gates. If a dependency or scope conflict remains, keep that unit pending and preserve independent work.
+
+Apply the [operating policy](../POLICY.md) and [project guide](../PROJECT.md). Record reusable decisions and improvements at the appropriate checkpoint.

package/dist/bun/native/payload/harness/workflows/W18.md

@@ -0,0 +1,13 @@
+# W18 Implement, resume and hand off
+
+**Trigger**. Implement this change, resume it, hand it off or tell me what is blocked.
+
+**Authority**. Use the accepted W16 definition, scoped work authorization and current project rules. Keep Git writes, publication and other effects independent.
+
+1. Verify the current contract, exact checkout, affected paths, last checkpoint and next action. Reconcile uncertain earlier effects from actual evidence before replaying anything.
+2. Work within the authorized scope. Read relevant source maps, use established tools, maintain tests and update durable checkpoints for decisions, progress, errors and next actions.
+3. Run W23 checks for applicable code and evidence. Obtain independent technical audit and separate independent code review. Update consumers, instructions and knowledge within scope before handoff.
+
+**Result and recovery**. Return what changed, verification and any pending gates with owner and resume condition. If the contract or authorization has changed, isolate the difference and route it through W16.
+
+Apply the [operating policy](../POLICY.md) and [project guide](../PROJECT.md). Record reusable decisions and improvements at the appropriate checkpoint.

package/dist/bun/native/payload/harness/workflows/W19.md

@@ -0,0 +1,13 @@
+# W19 Bounded investigation
+
+**Trigger**. Understand this code, audit sources or compare options.
+
+**Authority**. Investigate within approved evidence and project scope. Research access, external writes and implementation changes are separate.
+
+1. State the question, expected source inventory and what would resolve it. Use the project's index and source maps before opening larger artifacts.
+2. Fetch, filter and normalize with reliable tools. Read exact symbols or excerpts where needed, account for omissions and distinguish observation, inference and proposal.
+3. Reconcile the actual sources, revisions, pages, failed queries and omissions against the inventory. Compare findings against reproducible evidence. Use independent review for technical claims and preserve contradictions with the source role and unresolved scope.
+
+**Result and recovery**. Return the smallest useful analysis with locators and remaining uncertainty. Preserve a bounded evidence pointer instead of copying private raw outputs.
+
+Apply the [operating policy](../POLICY.md) and [project guide](../PROJECT.md). Record reusable decisions and improvements at the appropriate checkpoint.

package/dist/bun/native/payload/harness/workflows/W20.md

@@ -0,0 +1,13 @@
+# W20 Harness adoption and update
+
+**Trigger**. Set up, adopt or update the harness, or add a repository.
+
+**Authority**. Inspect locally and prepare a plan. Apply only the confirmed locations, modes, host entry points and services. Version control and external publication permissions remain separate.
+
+1. Inventory the installation root, registered repositories, applicable instructions, native tools and candidate services. Establish the workspace mode, per-repository versioning and whether budgets or limits apply.
+2. Use the installer to preview a new install, repo registration or update. For each capability record adopt, adapt, defer or reject, with authority and destination. Check that local-only paths use effective repo-local exclusions and that optional services and snapshots have explicit selections.
+3. Apply the authorized plan and verify the receipt, generated paths, exclusions and preserved local settings. Open a fresh selected host, invoke its setup entry and verify a representative task resolves the right workflow and project authority.
+
+**Result and recovery**. Return the installed version, enabled and pending capabilities, verified discovery and recovery steps. When drift or an access gap blocks a step, preserve the checkpoint and resolve the specific conflict.
+
+Apply the [operating policy](../POLICY.md) and [project guide](../PROJECT.md). Record reusable decisions and improvements at the appropriate checkpoint.

package/dist/bun/native/payload/harness/workflows/W21.md

@@ -0,0 +1,15 @@
+# W21 Source maps and knowledge indexes
+
+**Trigger**. Update source maps, indexes or project knowledge.
+
+**Authority**. Prepare or edit scoped local knowledge records. Keep raw sources and private output outside the portable package.
+
+1. Identify the changed implementation, authoritative sources, existing map threshold and intended consumers. Check current locators before generating new ones.
+2. Update relevant sibling maps, directory routers and project indexes with stable symbols or headings and small reading ranges. Capture decisions and scope without copying unrelated content.
+3. Verify generated ranges, links and exact navigation in a fresh session. Inspect the actual package or build inventory for developer-only material when the change affects distribution.
+
+**Result and recovery**. Return updated locations and verification. Preserve owner and revalidation trigger where a locator or authoritative link is unresolved.
+
+For a registered Git repository configured to keep source maps local, preview the exclusion from the installation root with `rtk proxy bun --no-env-file llm/harness/runtime/harness.mjs protect src/example.llm.md --project example --plan`. Use the actual registered repository and a path relative to that repository. Apply the confirmed scope with `--yes`. Inspect any tracked path conflict and keep it pending until the appropriate authorization resolves it.
+
+Apply the [operating policy](../POLICY.md) and [project guide](../PROJECT.md). Record reusable decisions and improvements at the appropriate checkpoint.

package/dist/bun/native/payload/harness/workflows/W22.md

@@ -0,0 +1,13 @@
+# W22 Reusable automation and maintenance
+
+**Trigger**. Automate this repeated operation or prevent a recurring failure.
+
+**Authority**. Use authorized local evidence and preserve the existing execution boundaries. External side effects remain separately scoped.
+
+1. Record the recurring operation, error or context cost and verify its cause. Find the existing native route, tool catalog and contracts before creating another utility.
+2. Reuse or extend the smallest reliable tool. Define explicit inputs, outputs, permissions, side effects, limits and original exit behavior. Keep a bounded raw debugging route and make retry safe where possible.
+3. Test expected, failure, recovery and edge cases. Run the changed workflow against representative evidence, update its catalog and affected consumers and confirm fresh-session discovery.
+
+**Result and recovery**. Return the verified improvement and regression protection. When judgment remains, document an example, counterexample and review gate.
+
+Apply the [operating policy](../POLICY.md) and [project guide](../PROJECT.md). Record reusable decisions and improvements at the appropriate checkpoint.

package/dist/bun/native/payload/harness/workflows/W23.md

@@ -0,0 +1,21 @@
+# W23 Verification and coverage
+
+**Trigger**. Verify this change, coverage or source policy.
+
+**Authority**. Run approved local checks. Launching remote jobs, changing policies and rewriting golden evidence require their own applicable permission.
+
+1. Identify the exact revision, source families, case denominator, applicable stack and established gates. Derive the denominator from the full relevant source inventory, including files no test loaded. For greenfield work, require measurable full coverage for every family and keep uninstrumented scope pending.
+2. Run the native format, lint, type and behavior checks through the project runner. Use the compatible personal style checks for authored TypeScript where the project lacks applicable Biome. Record actual selected, executed, failed, skipped and uncovered cases.
+3. Investigate failing tests and evidence before changing the test or implementation. Keep coverage distinct from deterministic style and independent factual audit and code review. Rerun affected checks after changes.
+
+**Result and recovery**. Return reproducible commands, measured scope and results. Flag empty runs, unresolved failures and unmeasured scope, with owner and resume condition.
+
+For LCOV evidence, independently inventory the applicable source files and write a scope file with workspace-relative paths and the required metrics. Use the metrics the project's tooling is expected to measure.
+
+```json
+{"files":["src/example.ts"],"metrics":["lines","functions","branches"]}
+```
+
+Run `rtk proxy bun --no-env-file llm/harness/runtime/harness.mjs verify coverage coverage/lcov.info --config llm/harness/coverage.json` from the installation root, substituting the actual report and scope file. The verifier requires one complete record per expected source, matches declared totals to measured entries, and reports the actual hit and found counts. A selected metric with no measurable entries stays pending. Record the run and source revision separately to establish that the report is fresh.
+
+Apply the [operating policy](../POLICY.md) and [project guide](../PROJECT.md). Record reusable decisions and improvements at the appropriate checkpoint.

package/dist/bun/native/payload/harness/workflows/W24.md

@@ -0,0 +1,13 @@
+# W24 Design and visual evidence
+
+**Trigger**. Design or verify a visual change or export.
+
+**Authority**. Read the approved design and supplied visual evidence. Editing, propagating, replacing a golden artifact and publication are separately scoped.
+
+1. Resolve the design authority, exact saved revision, variants, states, accessibility needs and representative consumers. Locate the applicable native tool and permitted operations.
+2. Use a small canary when testing propagation across files or components. Capture before and after evidence, native properties, actual rendered behavior and relevant interactions or exports.
+3. Check affected consumers and verify the exact saved revision. Compare reproducible exports with accepted golden evidence where available. Keep a design proposal separate from an accepted or published design.
+
+**Result and recovery**. Return verified design state, visual evidence and affected scope. If the native tool cannot establish parity, preserve the gap and a reproducible checkpoint.
+
+Apply the [operating policy](../POLICY.md) and [project guide](../PROJECT.md). Record reusable decisions and improvements at the appropriate checkpoint.

package/dist/bun/native/payload/harness/workflows/W25.md

@@ -0,0 +1,13 @@
+# W25 Test design and execution
+
+**Trigger**. Design, generate, count or run tests.
+
+**Authority**. Prepare and run approved cases in the specified environment. CI queueing, real data writes and policy changes need the corresponding authority.
+
+1. Map the accepted behavior to manual, API, browser or data-derived scenarios. Assign stable case IDs. For tabular inputs, parse and validate required fields and record skipped rows.
+2. Resolve the native runner, exact selection, environment, fixture scope and pinned dependencies. Use an authorized browser or device profile where required and preserve the case-to-requirement link.
+3. Run the scoped cases and reconcile discovery, executed, passed, failed, skipped and retry counts. Inspect a missing or empty result before making a coverage claim.
+
+**Result and recovery**. Return case identities, command, environment and measured results. Keep unexecuted cases distinct from passes and use W26 for a failure.
+
+Apply the [operating policy](../POLICY.md) and [project guide](../PROJECT.md). Record reusable decisions and improvements at the appropriate checkpoint.

package/dist/bun/native/payload/harness/workflows/W26.md

@@ -0,0 +1,13 @@
+# W26 Test failure diagnosis
+
+**Trigger**. Diagnose or fix this test failure.
+
+**Authority**. Start with read-only diagnosis and an approved reproduction. Fix only authorized local scope. Creating a bug or automation ticket is separate.
+
+1. Identify the exact case, attempt, retries, revision, environment and failure signal. Read relevant original traces, reports and fixtures with a bounded format-aware tool.
+2. Compare the assertions, setup and observed behavior with the accepted contract. Reproduce where authorized. Classify the cause as test, product, environment or insufficient evidence.
+3. Correct the verified cause within scope and add a regression check. Rerun the relevant case and affected integration checks. Keep bug, fixture and pipeline actions separate.
+
+**Result and recovery**. Return the cause, evidence, correction and measured rerun. For insufficient evidence, preserve failed attempts and the next discriminating check.
+
+Apply the [operating policy](../POLICY.md) and [project guide](../PROJECT.md). Record reusable decisions and improvements at the appropriate checkpoint.

package/dist/bun/native/payload/harness/workflows/W27.md

@@ -0,0 +1,13 @@
+# W27 QA data, access and test framework
+
+**Trigger**. Prepare QA data, access or a shared test framework release.
+
+**Authority**. Treat fixture data, user roles, environments, publishing and consumer updates as distinct effects. Use approved test identities and isolated data.
+
+1. Resolve the exact target, identities, roles, runtime artifact and authorized test data recipe. Prefer disposable, scoped fixtures and inventory existing state before writing.
+2. Verify fixture state, roles and cleanup where applicable. For a framework release, inspect the exact artifact and its consumers, then record compatibility and the pins that need updating.
+3. Preview the permitted write or publication separately, execute only the approved operation and verify identity, version and behavior in the actual consumer.
+
+**Result and recovery**. Return verified fixture or artifact identities, exclusions and cleanup or rollback path. Keep any unresolved consumer or access gap visible.
+
+Apply the [operating policy](../POLICY.md) and [project guide](../PROJECT.md). Record reusable decisions and improvements at the appropriate checkpoint.

package/dist/bun/native/payload/harness/workflows/W28.md

@@ -0,0 +1,13 @@
+# W28 Environment, migration and recovery
+
+**Trigger**. Start or recover a local environment, migration or backup.
+
+**Authority**. Identify user-owned processes and mutable state before acting. Starting services, migrating, backing up and restoring each use the exact applicable permission.
+
+1. Resolve the checkout, service versions, resource owners, actual listeners, data store and target environment with finite read-only probes. Isolate temporary state and test credentials.
+2. Plan the approved operation and its recovery. For migrations, inspect applied history and target schema. For recovery, verify backup identity, byte integrity, format and destination before replay.
+3. Apply only the exact covered operation. Verify application readiness, migration state or restored data independently with the native tool. Preserve original evidence and failure causes.
+
+**Result and recovery**. Return actual state, verified recovery or the precise blocker. Continue independent work if the target or owner remains uncertain.
+
+Apply the [operating policy](../POLICY.md) and [project guide](../PROJECT.md). Record reusable decisions and improvements at the appropriate checkpoint.

package/dist/bun/native/payload/harness/workflows/W29.md

@@ -0,0 +1,13 @@
+# W29 Incident, security and performance triage
+
+**Trigger**. Triage a live incident, security issue or performance problem.
+
+**Authority**. Collect scoped evidence from authorized live or supplied sources. Remediation, traffic changes and outward notices require their own authority.
+
+1. Establish actual serving revision, service, environment, time window and affected behavior. Preserve source locators and state when evidence is sampled, delayed or incomplete.
+2. Use targeted logs, traces or metrics. Define baseline, comparison inputs and measurement method. Verify a claimed vulnerability or cause against representative evidence and applicable versions.
+3. Separate diagnosis, mitigation and durable correction. Propose the smallest next action with verification and rollback. Independently review material technical claims.
+
+**Result and recovery**. Return current impact, evidence, gaps and authorized next steps. Keep a hypothesis pending until a discriminating check confirms it.
+
+Apply the [operating policy](../POLICY.md) and [project guide](../PROJECT.md). Record reusable decisions and improvements at the appropriate checkpoint.

package/dist/bun/native/payload/harness/workflows/W30.md

@@ -0,0 +1,13 @@
+# W30 Release, deployment and rollback
+
+**Trigger**. Prepare, verify or roll back a release.
+
+**Authority**. Resolve preparation, artifact publication, deployment, activation and rollback as separate effects. Obtain the exact applicable grant.
+
+1. Pin the artifact hash, version, target environment, configuration and expected state. Verify applicable checks, signing or provenance and rollback prerequisites.
+2. Preview the exact operation and impact. Run the authorized native route once, then verify actual saved artifact or serving revision and downstream readiness. Distinguish queued from completed work.
+3. If verification fails, preserve observed state and use the approved rollback plan only where its scope covers the exact target. Verify actual recovered state before reporting completion.
+
+**Result and recovery**. Return exact artifact, activation or rollback evidence. Keep preparation and unresolved live state explicit.
+
+Apply the [operating policy](../POLICY.md) and [project guide](../PROJECT.md). Record reusable decisions and improvements at the appropriate checkpoint.

package/dist/bun/native/payload/harness/workflows/W31.md

@@ -0,0 +1,13 @@
+# W31 Cost, context and model qualification
+
+**Trigger**. Check cost, tune context or qualify a model.
+
+**Authority**. Use approved telemetry, profiles and budget. Provider calls, paid evaluations and configuration changes require applicable scope.
+
+1. Resolve the model, runtime, tools, task, budget, data set and measurement method. Fix comparable inputs, versions and criteria before comparing results.
+2. Collect measured input, output, cache, workers and reserve accounting with deterministic tools where available. Separate provider-reported values, offline estimates and unavailable values.
+3. Compare quality, cost and latency against the accepted criteria. Verify source dates, units and inference boundaries independently. Keep budget and provider effects explicit.
+
+**Result and recovery**. Return comparable results with denominators and coverage. If telemetry or a provider route is unavailable, label the estimate and keep qualification pending.
+
+Apply the [operating policy](../POLICY.md) and [project guide](../PROJECT.md). Record reusable decisions and improvements at the appropriate checkpoint.

package/dist/bun/native/payload/harness/workflows/W32.md

@@ -0,0 +1,13 @@
+# W32 Application-agent workflow and learning
+
+**Trigger**. Build an application-agent workflow or learn from review corrections.
+
+**Authority**. Use the accepted implementation definition and permission model. Changes to stored state, prompts, provider configuration and outward effects require their corresponding authority.
+
+1. Define input, tool permissions, expected output, durable state and authorization boundaries. Combine scoped policy and task instructions with deterministic preparation and explicit effect previews.
+2. Implement recovery with exact checkpoints, correlation, failure handling and readback. Keep private tool payloads in the approved store and verify who owns human edits and review decisions.
+3. Capture confirmed corrections with provenance, scope, supersession and retirement rules. Route accepted actions to an authorized local list. Test repeat, retry and scope isolation and update discovery.
+
+**Result and recovery**. Return working behavior and evidence for each state transition, or a design clearly marked pending. Preserve verified lessons and distinct human decisions.
+
+Apply the [operating policy](../POLICY.md) and [project guide](../PROJECT.md). Record reusable decisions and improvements at the appropriate checkpoint.

package/dist/bun/native/payload/harness/workflows/W33.md

@@ -0,0 +1,13 @@
+# W33 Reusable prompt forms
+
+**Trigger**. Choose or design a reusable prompt.
+
+**Authority**. Prepare a prompt and bounded inputs for the current task. Running tools, delegating and external effects still follow their own permissions.
+
+1. Choose the request form that best fits the task. Use examples for input and output behavior, plan-first for unresolved requirements, constrained-source for bounded evidence, and compare-and-decide for explicit tradeoffs.
+2. State the target, output, authority, allowed sources and tools, constraints, evidence and completion checks. Mark unresolved choices and required capabilities explicitly.
+3. Test the prompt on a representative case, inspect where it overreaches or leaves gaps, and catalog a verified improvement with the intended trigger and limits.
+
+**Result and recovery**. Return the reusable prompt and a concrete example. If inputs or permissions are missing, preserve the form and leave dependent execution pending.
+
+Apply the [operating policy](../POLICY.md) and [project guide](../PROJECT.md). Record reusable decisions and improvements at the appropriate checkpoint.

package/dist/bun/native/payload/harness/workflows/W34.md

@@ -0,0 +1,13 @@
+# W34 Application changes and integrations
+
+**Trigger**. Implement a route, API field, component or external integration.
+
+**Authority**. Use the accepted W16 scope. Discover uncertain external contracts read-only. Live mutations, schema changes and release effects need the appropriate grant.
+
+1. Locate current architecture, authoritative contract, generated boundaries and exact application path. Follow data, authentication and authorization through each applicable layer.
+2. Inspect local examples and native interfaces. Verify required external fields, capabilities, environments and failure modes before coding. Resolve material discrepancies through W16.
+3. Implement within the approved boundary. Exercise the relevant interface, domain, persistence and consumer behavior, including failure paths. Run W23 and preserve any migration or release prerequisites.
+
+**Result and recovery**. Return exact code, contract and verification evidence. Keep deployment, unresolved integration behavior and out-of-scope layers visible.
+
+Apply the [operating policy](../POLICY.md) and [project guide](../PROJECT.md). Record reusable decisions and improvements at the appropriate checkpoint.

package/dist/bun/native/payload/harness/workflows/W35.md

@@ -0,0 +1,13 @@
+# W35 Test traceability and QA documentation
+
+**Trigger**. Update test traceability or sync QA documentation.
+
+**Authority**. Read the approved test and documentation sources and prepare a scoped diff. Tracker field edits, document creation and posting have separate effects.
+
+1. Resolve the canonical test and issue IDs, document section and version. Verify the project's artifact routing, native fields and update rules.
+2. Prepare the smallest exact patch. Validate identifiers, links, required fields, headings and current content and avoid overwriting unrelated edits.
+3. Preview each requested effect and apply only the approved changes. Read back exact persisted fields or section text and verify related links or index entries.
+
+**Result and recovery**. Return the approved diff or exact verified revision. Report version conflicts and destinations that transformed or rejected a submitted payload.
+
+Apply the [operating policy](../POLICY.md) and [project guide](../PROJECT.md). Record reusable decisions and improvements at the appropriate checkpoint.

package/dist/bun/native/payload/harness/workflows/W36.md

@@ -0,0 +1,13 @@
+# W36 Service identity and shared evidence
+
+**Trigger**. Verify a service identity, change a shared file or manage an evidence store.
+
+**Authority**. Resolve identity and capability read-only. Shared edits, resets, repository registration and cache changes have separate scope and permission.
+
+1. Verify the service, account, exact resource, environment, expected revision and allowed operation. Confirm credential references without reading or exposing secrets unnecessarily.
+2. Prepare exact changes and confirm concurrency and recovery controls such as conditional writes or supported history. Present payload, target and expected result before a write.
+3. Apply the authorized operation with the native protection and read back the exact result. Reconcile unexpected state or partial failure before any retry. Preserve earlier versions and audit evidence where applicable.
+
+**Result and recovery**. Return identity and revision-bound proof. If the route cannot guard concurrent changes or verify the required outcome, keep the effect pending and offer a local draft.
+
+Apply the [operating policy](../POLICY.md) and [project guide](../PROJECT.md). Record reusable decisions and improvements at the appropriate checkpoint.

package/dist/bun/native/payload/harness/workflows/W37.md

@@ -0,0 +1,13 @@
+# W37 Scheduler and queue work
+
+**Trigger**. Add or diagnose a scheduled job or queue worker.
+
+**Authority**. Inspect scoped scheduler, producer and consumer evidence first. Registration, replay and downstream writes each require their own authority.
+
+1. Resolve schedule, tenant, producer, queue, consumer, payload references and actual serving revision. Collect bounded structured evidence for timing, delivery, attempts and results.
+2. Distinguish scheduler registration, enqueue, delivery, processing and downstream effects. Verify retry, deduplication and isolation behavior before considering a replay or local fix.
+3. Implement or replay only the authorized scope. Verify the actual event and resulting state separately at each applicable boundary. Preserve correlation and failure evidence.
+
+**Result and recovery**. Return the exact job or event, observed states and next step. If an effect remains uncertain, preserve the checkpoint before retrying to avoid duplicates.
+
+Apply the [operating policy](../POLICY.md) and [project guide](../PROJECT.md). Record reusable decisions and improvements at the appropriate checkpoint.

package/dist/bun/npm/LICENSE

@@ -0,0 +1,21 @@
+MIT License
+
+Copyright (c) 2026 AppToolkit contributors
+
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in all
+copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+SOFTWARE.

package/dist/bun/npm/README.md

@@ -0,0 +1,33 @@
+# AppToolkit Harness
+
+AppToolkit Harness installs project guidance, deterministic verification tools and 37 reusable engineering workflows for Codex, Claude Code and OpenCode. It supports new projects, existing repositories and a shared harness for related repositories.
+
+Use macOS or Linux with Bun 1.4.2 or later, or Node.js 24 or later with npm. Use Git for versioned projects and RTK for project tooling. Run the installer from the target directory using one of these commands.
+
+```sh
+rtk proxy bunx --bun @apptoolkit/harness install
+rtk proxy npx @apptoolkit/harness install
+```
+
+`bunx --bun` selects Bun even when Node.js is absent. Plain `bunx` also works when Node.js is on `PATH`.
+
+Review the proposed files, then select which agents, repositories and optional services to configure. Open the same directory in your agent and run `$apptoolkit init` in Codex or `/apptoolkit init` in Claude Code or OpenCode. The installed `llm/HARNESS.md` is the shared entry point.
+
+In a directory of related repositories, start the agent in the parent directory. The central harness keeps shared decisions and scoped repository context. Use `$apptoolkit add repo <path>` or `/apptoolkit add repo <path>` after cloning another repository. Repositories can keep their own guidance and toolchain.
+
+To preview an installation in a noninteractive session, inspect the target and supply the explicit choices.
+
+```sh
+rtk proxy bunx --bun @apptoolkit/harness inspect --root /absolute/workspace --json
+rtk proxy bunx --bun @apptoolkit/harness install --root /absolute/workspace --hosts codex,claude,opencode --local --repo-mode .=local --plan --json
+```
+
+Use `--yes` in place of `--plan` to apply a reviewed noninteractive plan. Select the workspace mode separately from each Git repository's generated files. Local mode uses the repository's `info/exclude` metadata for the planned files. Use `--repo-mode <path>=local|versioned` for each selected Git repository.
+
+Use `rtk proxy bunx --bun @apptoolkit/harness@latest update --plan --json` or `rtk proxy npx @apptoolkit/harness@latest update --plan --json` to preview a release update. The updater checks managed files against the installation receipt. Resolve reported conflicts before applying the update. Keep project decisions in the installed local profile, and review the affected paths and version before applying.
+
+Plane and Penpot are separate opt-ins. To make a provider available after installation, preview an update with `--enable-service plane` or `--enable-service penpot`, apply it, and complete the project connection through `init`. The installed runbooks explain instance setup, private credentials, authorization and readiness checks. Snapshot verification requires a versioned harness and project-level opt-in. Use Bun to install the isolated Biome dependency before first using the author style profile.
+
+The installed local CLI reads the workflow index, retrieves one workflow at a time and checks Markdown against a local document policy. See `llm/HARNESS.md` for the exact commands. The agent records which operations are configured and which were verified in the current environment.
+
+Distributed under the MIT license.