wiki-search

name	wiki-search
description	Use when finding, retrieving, and validating information in a Hermes memory wiki vault before answering or deciding whether indexes need refresh.
version	0.1.0
author	Nous Research
license	MIT
metadata	{"tools":["wiki_search","wiki_get","wiki_status","wiki_reindex"]}

Overview

Search the Hermes memory wiki with a read-first workflow. Use this skill when the answer may depend on stored wiki knowledge, prior project memory, source-backed claims, or compiled wiki indexes.

When to Use

• The user asks about stored wiki knowledge, prior work, project memory, entities, concepts, sources, reports, or syntheses.
• You need to locate source-backed claims or exact claim IDs before answering.
• You need a page path to cite or to feed into an authoring/maintenance workflow.
• Search returns too few results and recent edits may not be indexed yet.

Search Workflow

1. Run wiki_status if the vault path, page count, or readiness is uncertain.
2. Search first with wiki_search using a concise query and appropriate result limit.
3. Open promising results with wiki_get; snippets are discovery hints, not the authoritative page.
4. Cite wiki page paths and claim IDs when they support an answer, for example entities/ada.md or claim.ada.role.
5. Refine with names, aliases, page paths, source IDs, or claim terms from retrieved pages when needed.
6. If freshness matters after recent edits or a compiled index looks stale, run wiki_reindex, then repeat wiki_search.
7. Base answers on retrieved content and state uncertainty when pages, evidence, or claims are incomplete.

Pitfalls

• Do not answer from snippets alone when exact page text matters; use wiki_get.
• Do not run wiki_reindex repeatedly unless there is a freshness reason.
• Do not assume an empty result means no knowledge exists; try narrower and broader wiki_search queries.
• Confirm vault readiness with wiki_status when failures suggest a missing or uninitialized vault.

Verification Checklist

• wiki_status was used when vault readiness was unclear.
• Relevant records were discovered with wiki_search.
• Exact page content was checked with wiki_get before making final claims.
• Page paths and claim IDs were cited when relevant.
• wiki_reindex was used only when search freshness was a plausible issue.