skills.rest
growilabs
growilabsOFFICIAL
Explore all skills by this author →
GitHub

suggest-path-evaluator

Evaluates GROWI suggest-path save-path proposals with per-candidate plausibility verdicts.

1.5k|244|Updated Feb 16, 2017
Search Tags:#evaluation#growi#mongoDB#Wiki#batch aggregation#path suggestion#suggest-path
One-click install
npx skills add https://github.com/growilabs/growi --skill suggest-path-evaluator
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: suggest-path-evaluator
Source: https://github.com/growilabs/growi/tree/main/.claude/skills/suggest-path-evaluator
Command: npx skills add https://github.com/growilabs/growi --skill suggest-path-evaluator

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes scripts (resource) and references (resource) components.

What problem does it solve?

GROWI's suggest-path feature proposes parent paths for saving a document, but there is no systematic way to judge whether those proposals are actually usable. This Skill reviews each proposed path against the wiki's real content and organization, marking it usable, improvable, or not usable, so you can measure and tune suggest-path quality instead of eyeballing results.

Core Features & Use Cases

  • Per-candidate verdicts: Judges every suggest-path proposal as o (usable), triangle (usable but a better home exists), or x (not usable), using wiki-agnostic principles rather than a fixed answer key.
  • Drilldown to the ideal home: Descends the wiki tree (beam search, up to 3 children per level, depth cap 5) to build the document's recommended save location, which backs x verdicts and supplies the better place for triangle verdicts.
  • Batch aggregation and consistency gating: Ships scripts that reconcile per-document digests against bodies to catch body swaps, and aggregate verdicts into per-domain usable-rate, per-candidate rates, recall-miss rate, and x-by-reason breakdowns.
  • Use Case: While dogfooding suggest-path on a devcontainer GROWI wiki, run the evaluator on a batch of documents to find which wiki areas produce the most unusable path suggestions and why.

Quick Start

Ask the AI to evaluate suggest-path's proposals for a document body against the local devcontainer GROWI wiki and report the per-candidate verdicts plus the recommended home.

Frequently Asked Questions about suggest-path-evaluator

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I evaluate GROWI suggest-path results for a document?▼

Provide the document body and the target GROWI wiki; the skill calls suggest-path via the MCP tool or the HTTP endpoint /_api/v3/ai-tools/suggest-path, then judges each proposed path as usable, improvable, or not usable. It also drills down the wiki tree to report the document's recommended home.

What does the suggest-path evaluator use as the correct answer?▼

It uses no single answer key. Each candidate is judged on plausibility against the wiki's own organization, and multiple candidates can be marked usable. A drilldown-built recommended home acts only as a reference for spotting clearly better placements.

How does the evaluator read the GROWI wiki tree and page content?▼

It clones the wiki through GROWI Vault, which exposes the wiki as a bare git repository of Markdown files, then uses ls, read, and grep on the clone. A MongoDB direct-read fallback against the pages and revisions collections is available when Vault is unavailable.

Can I run the suggest-path evaluation on a batch of documents?▼

Yes. Run the flow per document, then use scripts/reconcile-digests.ts to verify each digest matches its body and scripts/aggregate.ts to compute per-domain usable rates, per-candidate verdict rates, recall-miss rate, and an x-by-reason breakdown.

What are the limitations of the suggest-path evaluator?▼

It is a measurement tool and does not modify suggest-path or GROWI. Its usable-rate is precision-side only and cannot detect good locations suggest-path never proposed, and judgments made against a partial or stale tree can overstate misses.

Related Skills

Explore More Software Engineering Skills→
samuelcastrosamuelcastro

wiki-setup

Create an Obsidian wiki vault with folders, starter files, and configuration templates.

Community
Intermediate
MantisWareMantisWare

self-consistency

Generate multiple reasoning paths and aggregate final answers by majority voting.

Community
Advanced
FolahanWilliamsFolahanWilliams

noiseaudit

Compute standard deviation across sub-agent judgments to surface consensus gaps.

Community
Advanced
anthropicsanthropics
9.1k

tabular-review

Extract schema-based contract data into spreadsheet grids with source citations.

Official
Advanced