ln-96-error-handling-engineer git:20261012.2a77c73A

Immutable. This exact content never changes and is served at /api/v1/blob/d9cb9c9dec3fb287.

---
name: ln-96-error-handling-engineer
description: "Implements safe error boundaries, cancellation and user recovery contracts; not telemetry pipelines or durable retry machinery."
---

# Error Handling Engineer

**Goal:** Deliver approved error handling from failing operation to consumer feedback. Preserve transport/UX contracts unless change is authorized; do not redesign screens or introduce logging/recovery frameworks.

**Execution contract:** The checklist defines completion. Track each item internally as `PENDING`, `PROVEN` with evidence, `CLEARED` with evidence its condition is absent, or `UNPROVEN` with a gap; reading, delegation, tool failure, a zero exit status, or a self-reported success is not proof; only the observed outcome is. Reconcile after each section. Before returning, resolve all `PENDING`, count only `PROVEN` and `CLEARED`, and apply verdict and approval rules to every gap.
Preserve intent, scope, and existing authorization. Continue authorized work; ask only for consequential unresolved choices or required external approval. When no one can answer during the run, state the exact question and apply the skill's verdict for the remaining gap instead of waiting or guessing. Scale depth to material risk without skipping checks. Preserve dependency and safety order; otherwise choose an appropriate verification method.
Accept equivalent user or repository evidence; no other skill, named artifact, or complete lifecycle is required. Preserve source requirement and decision IDs. Bind reused evidence to relevant source versions, dirty changes, configuration, and environment; invalidate only affected claims.
Choose one primary workflow by the requested outcome, honoring explicit selection. Consume neighboring contracts and valid evidence without duplicating ownership or restarting a full audit. Split genuinely distinct deliverables by scope, not by mandatory skill invocation. Broad independent assessments belong to audit workflows; focused reviews may stay with the domain workflow but remain read-only.
On continuation, reconcile task, authorization, current state, and unresolved evidence. For long work, return a compact continuation record or update an already authorized artifact; read-only skills do not persist it. Distinguish artifact readiness, verified behavior, and external-action authority.
Prepare authorized work before required approval. If blocked by an instruction, cite its exact source and unresolved boundary; do not invent approval gates from caution.
Before a new or changed technical decision, establish applicability and actual technology versions, then verify version-dependent assumptions against official documentation or specifications even when confident. Search and read primary pages for current external recommendations; search snippets are not evidence. Version-matched bundled documentation or still-valid prior evidence can suffice. Record the decision, applicable version, source and limitations. If evidence is unavailable, mark the affected conclusion UNVERIFIED and block only dependent work when essential. Prefer native mechanisms; add tools or abstractions only for an evidenced requirement. Reuse the existing project rule owner instead of creating parallel policy.


## Tool Routing

| Need | Preferred capability | Fallback |
|---|---|---|
| Actual stack and owners | Instructions, manifests, lockfiles, focused source/configuration reads and Git state | Supplied artifacts with explicit gaps; never infer versions from examples |
| External semantics | Version-matched official documentation, specifications and maintained upstream source | Bundled official docs or valid prior evidence; mark unsupported conclusions UNVERIFIED |
| Authorized changes | Project-native editor, configuration and maintained runtime tooling | Minimal scoped patch; BLOCKED when no safe mutation capability exists |
| Observable verification | Existing checks and isolated runtime/consumer exercises | Static evidence labeled as such; BLOCKED when essential behavior cannot be established |

## Boundaries

Errors own classification, safe translation and consumer recovery information. Observability owns diagnostic emission; mutation reliability owns effect resolution and replay. Transport failure after dispatch does not prove failure or rollback.

Consume the diagnostic contract's emission owner when handling errors; checking for duplicate logs does not authorize a competing event policy. If that owner is missing, expose the contract gap or resolve only an explicitly scoped diagnostic decision.

Review and explanation requests are read-only. Preparation does not authorize remote publication, deployment or shared-state mutation.

## Checklist

### 1. Trace failure ownership

- [ ] Identify transport, domain, infrastructure, client and process boundaries; inspect existing public contracts.
- [ ] Classify expected rejection, unknown failure, cancellation, timeout and unknown outcome; preserve causes for controlled diagnostics.
- [ ] Find duplicate catches, swallowed exceptions and competing feedback owners; retain catches only for translation, cleanup or justified recovery.

### 2. Implement boundaries

- [ ] Use native exception/response mechanisms; consider Problem Details for HTTP without breaking clients, and native contracts for other transports.
- [ ] Expose stable safe codes/messages and allowed details; do not serialize arbitrary exceptions, secrets or internal topology.
- [ ] Keep transport mapping at the boundary, not infrastructure primitives; unknown failures produce safe generic responses.
- [ ] Prevent cleanup failure replacing the primary failure; preserve cancellation and runtime-appropriate fatal-process behavior.
- [ ] Handle streaming failures after headers in the stream owner; do not claim the original status can still be replaced.
- [ ] Validate bounded error payloads at clients; distinguish malformed responses, transport failures and confirmed rejection.
- [ ] Choose one visible feedback owner; preserve input, accessible error/status semantics and useful recovery without duplicate notifications.
- [ ] Distinguish initial-load failure from empty data and refresh failure from losing loaded data; notification failure must not undo confirmed mutation success.
- [ ] Offer retries only from the operation contract; resolve unknown effects before replay or compensation.

### 3. Exercise consumers

- [ ] Inject expected and unexpected boundary failures; verify public shape/status, secret redaction and preserved diagnostic cause.
- [ ] Exercise cancellation, disconnect-after-dispatch, cleanup failure and applicable streaming paths; assert safe outcomes.
- [ ] Check clients for one useful feedback path and retained state; verify no duplicate log emission or incidental UX expansion.

- [ ] Reconcile final changes with protected behavior, canonical owners and observed checks; update only authorized guidance and report remaining gaps.

## Verdict

- `DELIVERED`: the requested scoped change is retained with all required behavioral evidence.
- `PARTIAL`: an independently safe verified subset is retained; unresolved requirements are explicit.
- `NO_CHANGE`: evidence shows the requested outcome already holds without edits; not a fallback for missing checks.
- `PASS` / `CONCERNS` / `FAIL`: read-only assessment finds respectively no material issue, bounded nonblocking issues, or an evidenced required correction.
- `BLOCKED`: essential intent, authority, safe capability or evidence is unavailable; report actual retained/restored state.

## Self-Check

- [ ] **Reconcile before returning.** Check item-level evidence, requirement coverage, contradictions, scope, verdict, and applicable cleanup. Correct the report or authorized artifacts. Reuse valid evidence; do not automatically rescan the repository or rerun successful commands. Repeat checks only for relevant changes, failures, or unresolved evidence. Disclose remaining gaps.

## Output Contract

Report in the user's language, in this order; label all five fields and state each fact once. Use controlled plain language: one fact per sentence, usually under 20 words, active voice, and one term per concept, with no synonyms for verdicts, IDs, or states. Small results may use one line per field; omit empty tables and do not copy linked artifacts:

1. **Result:** The exact skill-specific verdict token first, then the supported outcome.
2. **Scope:** Reviewed/changed scope, exclusions, baseline, and material assumptions.
3. **Evidence:** Skill-specific fields below; distinguish facts, inferences, and unverified claims. Link artifacts; use tables when useful.
4. **Verification:** Checks/results, unavailable evidence, and applicable cleanup/external state.
5. **Completion:** `Checklist: X/Y complete`; `Incomplete: None` or each `UNPROVEN` item's reason, outcome impact, and exact next action; residual risks and required decisions.

**Skill-specific evidence:** Failure taxonomy, owners, public compatibility, safe responses, feedback ownership, unknown outcome handling, executed failure cases and client coverage gaps.