인터페이스는 현지화되어 있으며, Skill 소개와 설명 문서는 선택한 언어로 AI가 자동 번역합니다. 코드·명령어·경로·전문 용어는 원문을 유지하며 원작자 문서가 우선합니다.
Skills 둘러보기
생산성 및 협업 · SKILL.md

lucidchart-common-errors

lucidchart-common-errors

Triage Lucid authentication, scope, version, import, export, extension, and connector failures from bounded evidence. Use when a Lucid integration fails. Trigger with "troubleshoot Lucid".

Independent GitHub snapshot · Automated classification · Original documentation governs

Skill 문서

GitHub

name: lucidchart-common-errors description: 'Triage Lucid authentication, scope, version, import, export, extension, and connector failures from bounded evidence. Use when a Lucid integration fails. Trigger with "troubleshoot Lucid".' argument-hint: "[symptom] [integration-kind]" allowed-tools: Read, Glob, Grep, WebFetch, Write, Edit version: 1.8.0 author: Jeremy Longshore jeremy@intentsolutions.io license: MIT tags: [saas, lucidchart, troubleshooting, oauth, diagnostics] model: inherit effort: medium compatibility: Designed for Claude Code; live diagnosis requires authorized Lucid evidence and approval from the document, application, connector, or account owner

Lucid Integration Error Triage

Overview

Identify the failing layer before changing credentials, scopes, documents, packages, or connector state.

Prerequisites

  • A redacted request ID, timestamp, status, integration kind, endpoint or command, and API/package version
  • Expected token type and scopes from the exact official operation page
  • Permission to inspect the relevant repository and Lucid project

Tool Discipline

Use Read, Glob, and Grep for local evidence, WebFetch for official Lucid contracts and status, and Write or Edit only for approved fixes or redacted receipts.

Current Contract

  • 401 can reflect invalid or revoked credentials; 403 can reflect token type, scope, resource access, or endpoint rules.
  • 429 behavior and limits are endpoint-specific; never substitute a pack-wide quota.
  • Standard Import, REST resources, editor extensions, and data connectors have different failure surfaces.

Authentication

Never collect raw tokens, secrets, authorization codes, cookies, or complete request headers. Confirm only credential class, grant, scope names, issuer environment, rotation state, and a redacted fingerprint.

Instructions

  1. Establish the last known good time and whether Lucid status reports an incident.
  2. Classify the layer: network, auth, authorization, version, payload, document permission, extension build, or connector callback.
  3. Compare token type, scopes, endpoint, method, content type, and Lucid-Api-Version with the exact current operation page.
  4. For imports, inspect archive paths, document.json, IDs, product, size, and media references without uploading customer content.
  5. For extensions, inspect manifest scopes, package versions, build output, developer mode, and browser console evidence.
  6. Reproduce with the smallest sanitized fixture and one reversible action.
  7. Record the proven cause, fix, verification, rollback, and evidence gaps.

Approval Boundaries

Do not broaden scopes, rotate credentials, alter sharing, upload documents, replay callbacks, or publish packages without owner approval.

Output

Return symptom, layer, evidence, ruled-out causes, confirmed cause, bounded fix, verification, rollback, and owner actions.

Error Handling

ConditionResponse
Evidence contains a secretStop, redact, rotate through the owner, and do not retain it.
Endpoint contract differs from memoryTreat current official documentation as authoritative.
Cause remains ambiguousReturn competing hypotheses and the smallest discriminating test.

Example

layer=authorization; status=403; token=user; expected-scope=lucidchart.document.content:readonly; secret-retained=no; next=owner-reconsent

Resources

Next Steps

Turn a confirmed recurring failure into a sanitized regression test.

저장한 Skills