SuperAPI AI Directory

lazyweb-explain-flow

Agents & models

View repository

Updated 10-15 23:49

About this skill

Lazyweb — Explain a Flow

Host an explanatory diagram: a sectioned flow-view page, stored insert-only — every call hosts a new standalone page that never collides with anything else.

MCP Setup

Use the hosted Lazyweb MCP tools at https://www.lazyweb.com/mcp.

Required tools:

  • lazyweb_health — verify connectivity when the MCP surface is uncertain
  • lazyweb_explain_flow — save the explanatory diagram (returns explainer_id + explainer_url)

If Lazyweb MCP is missing — or lazyweb_explain_flow isn't in the tool list — tell the user to run curl -fsSL https://www.lazyweb.com/install.sh | bash, reload, rerun.

MCP plan responses

Inspect every data-bearing tool result before applying its normal schema:

  • MCP_PRO_REQUIRED: relay the server message and returned intent-bound upgrade_url to the user, then stop. Do not retry another data tool or fall back to web/manual output.
  • FREE_REPORT_DAILY_LIMIT: relay the server message and returned intent-bound upgrade_url to the user, then stop. Do not retry another data tool or fall back to web/manual output.
  • Successful status: "locked_preview": relay display_to_user verbatim (or the returned MCP text if that is all the client exposes), including the preview and upgrade links. It is terminal: do not poll, retry another data tool, reconstruct output, or fall back.

Steps

  1. Understand what needs explaining. A question ("how does search work?"), a failure ("why did this report ignore the intent?"), a walkthrough, a comparison. Ground it in the real code/logs — read the handlers, don't invent.
  2. Build the explanatory diagram in the sectioned flow-view schema (sections, actors, nodes/edges with notes and real data payloads, bidirectional on round trips; an explanation earns its keep with REAL content, not boxes). Data snippets meet an exactness bar: actual payloads from the code/logs, … only to redact secrets (never to shorten content), and a sliced array keeps its first item(s) full-shape with the omission count in the note. Shape it for the QUESTION: order sections as the narrative, name the section titles as claims (e.g. "Where the run died — no model call"), put the "so what" in each node's note.
  3. Save it: lazyweb_explain_flow({ title, summary, product?, diagram }) → share the returned explainer_url.

When NOT to use

  • Proposing changes the user should Accept/Decline → /lazyweb-propose-ui-changes.

Notes

  • Explainers are point-in-time and cheap — make as many as the conversation needs; they never collide with each other.