Skip to main content

fallback-execution

API 调用失败时的强制降级执行:切换 skill 并通过浏览器/RPA/人工兜底完成任务

Quellinformationen

Repository
alextiannus/git-plugin-amc
Letzte Quellaktivität
31. Mai 2026 um 08:28
Erkannte Sprache von SKILL.md
Englisch
Sterne
0
Forks
0

Installationsoptionen

Standardmäßig ist der Prompt ausgewählt, der zuerst die Quelle prüft. Sie können zu einem direkten Befehl wechseln oder eine lokale Kopie herunterladen.

Quelldateien prüfen

Lesen Sie SKILL.md und alle von SkillsMP angezeigten Begleitdateien, bevor Sie sich für eine Installation entscheiden.

SKILL.md wird angezeigt

SKILL.md
Quellanweisungen · Schreibgeschützte Vorschau
name
fallback-execution
description
API 调用失败时的强制降级执行:切换 skill 并通过浏览器/RPA/人工兜底完成任务
plugin
git-plugin-amc
version
1.0.0
# Skill · Fallback Execution # F&B Content Engine · operations/fallback-execution ## When To Use Use this skill immediately when the primary API path cannot complete the task. Trigger conditions: - HTTP 4xx/5xx repeated failures after one retry - API capability not supported for target operation - OAuth/session/token invalid and not recoverable in current step - Platform-specific anti-abuse checks block direct API flow ## Mandatory Policy - API failure does not end the task. - The agent must switch to another executable path and continue. - Only mark task pending when all fallback paths are exhausted or human input is truly required. ## Fallback Order 1. Alternate skill in same plugin - Invoke platform execution skills first (for example `xiaohongshu`, `rednote`, `instagram`, `tiktok`). 2. Browser automation - Use browser-based workflow to complete publish/comment/reply/search actions. 3. RPA/script path - Use available CLI/script workflow if browser-first path is unstable. 4. Manual handoff package - Generate complete posting/comment package for human completion as last resort. ## Required Kanban Logging Every fallback attempt must be recorded in Kanban task description: - `primary_api`: endpoint/tool that failed - `failure_reason`: concise error cause - `fallback_path`: skill/browser/rpa/manual - `attempt_result`: success/partial/failed - `evidence`: post link, comment ID, screenshot note, or command output summary ## Execution Template Use this structure in task updates: ```text [FALLBACK] primary_api=<name> [FALLBACK] failure_reason=<reason> [FALLBACK] switched_to=<skill_or_method> [FALLBACK] result=<success|partial|failed> [FALLBACK] evidence=<url/id/log> ``` ## Practical Mapping - Publish API failed: - switch to platform skill + browser publish flow - Comment API failed: - switch to browser comment/reply flow - Search/detail API failed: - switch to crawler/browser extraction path - Media upload API failed: - switch to browser upload with verification step ## Stop Conditions Set task to `pending` only if one of these is true: - login/captcha requires human interaction - account permission missing and cannot be fixed automatically - content blocked by compliance policy and needs owner decision Otherwise continue execution until done.
Auf GitHub ansehen