AGENT SKILL

supportpages

Create or retry illustrated help articles with WTFM (by SupportPages.io) using its MCP connection and article writer. Use for WTFM or SupportPages.io article requests, including generating an unpublished draft for review.

License 许可
Apache-2.0
最近更新
2026-11-14
Stars
★ 1.6k
Forks
491

01 前言 Front Matter

name
supportpages
description
Create or retry illustrated help articles with WTFM (by SupportPages.io) using its MCP connection and article writer. Use for WTFM or SupportPages.io article requests, including generating an unpublished draft for review.

03 SKILL.md 正文

<!-- Managed by SupportPages: article coordination v1 -->

WTFM articles

Use the SupportPages.io MCP connection for the current project.

For inventory questions such as "what support articles do we have?", call supportpages_list_articles directly, without status or initialization first. When signed out or in local mode it reads the project's current local article files. Report those titles and paths without asking the user to sign in. An empty local list means no local articles were found, not that the hosted help centre is empty. Do not infer the inventory from Git status. Use source: "hosted" only when the user explicitly asks what is on SupportPages.io.

For article creation, check supportpages_status, finish initialization if required, and call supportpages_prepare_article before writing. Follow the returned readiness or capacity instructions. Show the prepared editor link; a preview is not a finished draft.

When the prepared run has prefer_background: true, delegate the writing task to a background subagent. This skill explicitly requests delegation for that article. Use the SupportPages.io writer and the returned Codex launch instructions, including the complete task brief, model settings and self-contained entrypoint. Pass the complete task brief directly in the subagent tool's task/message argument on launches and retries. Do not copy it into a temporary JSON file, create a handoff script, or print it with a shell command. The existing entrypoint already loads the saved run; no parent-side file edit is needed to launch the writer. If no custom-agent selector exists, use a named SupportPages.io subagent with the supplied writer instructions. The writer runs the entrypoint before generating anything, uses the command wrapper, and runs the finish command after validation.

Prioritize completion and approvals over quiet operation. After launch, record the actual execution mode and host task identity with supportpages_update_run. Show the editor link and say "launched" unless actual writer startup is confirmed. Keep the main turn active and supervise the writer using the host's agent wait/message facilities. Use bounded waits and concise progress updates; do not end the turn merely because launch succeeded.

The writer may send milestone updates and must surface blockers immediately. When an operation needs permission and host policy permits it, the writer should request approval directly through the host's native approval mechanism, with the exact command and reason. A sandbox restriction is not an explicit user denial. Do not suppress approval prompts or route them only to an idle parent. Continue after approval is granted. An explicit denial or a policy forbidding approval requests stops that action; never bypass it or enable YOLO.

If a background writer cannot surface approvals, confirm it has stopped and continue the same article in the foreground using MCP recovery instructions and the normal approval mechanism. Foreground execution retains the same permissions. If the host cannot supervise a delegated task, use foreground execution. Keep the MCP session alive; do not launch a separate agent CLI.

Inspect the writer's final result and saved run. A host "Completed" notification does not establish success. For a stopped writer that failed or is permission_blocked before generation finished, record event: "failed" and stopped: true with supportpages_update_run, then show the specific cause, recovery and editor link. Do not mark a writer failed while awaiting approval. If generation succeeded but delivery is pending or failed, recover with supportpages_complete_article without regenerating or marking the writer failed.

The main agent owns MCP calls. After successful writing, call supportpages_complete_article to confirm delivery, report the final editor link and follow its completion instructions. Leave the article unpublished until the user affirmatively asks to publish that draft.

Local folders. When supportpages_status or supportpages_init reports local, the folder has no help centre and needs no sign-in: articles are saved as Markdown with their screenshots. There is no capacity check, preview or editor link. Show the saved markdown path returned by supportpages_complete_article as the deliverable and never ask whether to publish automatically. Mention once that running wtfm publish in the terminal hosts the saved articles.

If the user explicitly asks to host or publish saved local articles, call supportpages_publish. Follow its browser approval and help-centre selection steps, then pass the selected article slugs. It uploads drafts and connects the folder for future articles. Do not accept credentials in chat or approve sign-in for the user. Making those drafts publicly visible is a separate action using supportpages_publish_article when authorized.

Hosted analysis and videos. Use supportpages_find_article_gaps, supportpages_suggest_sections, supportpages_recommend_articles or supportpages_create_video_walkthrough for the requested action. These tools check the prerequisites and open required browser setup pages. When signed out, first explain that an account/device sign-in is needed; sign in or create an account through the opened approval page. Do not begin with a repository invitation or send the user to the app homepage. Show the comparison code, and show the link as a fallback if browser_opened is false. Never approve browser consent.

After the user completes a step, call supportpages_setup_hosted with the returned next_arguments. Ask which help centre to use when project_required is returned (or use supportpages_create_project if the user wants a new one), then pass its project_id. The next required device-consent or repository setup page opens automatically. Preserve the original request and resume_arguments through these steps; do not start local analysis, recording or unrelated jobs. Once setup is available, retry resume_tool with resume_arguments. Walkthroughs require a completed hosted article: use supportpages_list_articles with source: "hosted" to select it. Do not restore deleted files or upload a local article without the user's explicit instruction. Finished videos remain private until explicitly shared.

Honor an explicit request to return immediately, explaining that approvals and completion may need another user turn on hosts that do not wake the parent. Retain the run ID and host task identity; reconcile the run on resumption.

Honor an explicit foreground request. If delegation cannot run, distinguish a missing host tool, an explicit session restriction, or the actual launch error; do not infer that background execution is disabled merely because the user did not separately request a subagent. Explain the specific limitation and use the same prepared brief in the foreground. Stop any launched writer before switching modes. Do not override host restrictions or launch a separate agent CLI.

For retries, follow the MCP recovery instructions and use supportpages_retry_article after stopping the old writer. If generation already finished and delivery failed, recover delivery instead of regenerating.

04 同仓库更多 Skills

hashgraph-online
hashgraph-online/awesome-codex-plugins

worktable-review-with-annotations

Create, inspect, reply to, update, or resolve situated Worktable annotations. Use when feedback belongs to a Doc section, block, text range, or HTML Doc, or when the user asks to address existing review feedback and coordinate a requested artifact edit.

★ 1.6k
hashgraph-online
hashgraph-online/awesome-codex-plugins

worktable-find-and-synthesize-context

Find, assess, and synthesize context from Worktable Docs and Records. Use when the user asks what a workspace contains, wants workspace-backed research, needs current facts reconciled across artifacts, or requests a cited synthesis without changing workspace state.

★ 1.6k
hashgraph-online
hashgraph-online/awesome-codex-plugins

worktable-create-or-update-docs

Create, edit, rename, or maintain narrative Worktable Docs. Use for plans, decisions, briefs, notes, research syntheses, and other durable prose when the user requests a new artifact or a safe update to an existing one.

★ 1.6k
hashgraph-online
hashgraph-online/awesome-codex-plugins

worktable-create-or-manage-records

Organize repeated structured information as durable Worktable Records. Use for research datasets, inventories, directories, pipelines, trackers, and other collections whose items need identity, fields, filtering, links, validation, or individual updates.

★ 1.6k
hashgraph-online
hashgraph-online/awesome-codex-plugins

worktable-create-or-manage-interactive-html

Create or improve durable visual and interactive artifacts in Worktable. Use for dashboards, trackers, calculators, mini apps, slide presentations, and other requests that benefit from a polished HTML Doc.

★ 1.6k
hashgraph-online
hashgraph-online/awesome-codex-plugins

worktable-collaborate-threads

Read, answer, and hand off messages through durable Worktable threads with users or other authenticated participants. Use when the user asks to check addressed messages, reply to a participant, continue a thread, request another agent's help, wait briefly for a response, or watch for comments and messages sent to the agent.

★ 1.6k