/maintain
Wiki maintenance — lint broken links, merge near-duplicates, upgrade confidence, flag stale pages via freshness tiers, gap analysis, concept synthesis. Use on: 'wiki maintenance', 'wiki cleanup', 'fix wiki', 'wiki health', 'check wiki', 'consolidate wiki'.
Claude Code
$ npx -y skills add Oshayr/LLM-Wiki --skill maintain --agent claude-code
Codex
$ npx skills add Oshayr/LLM-Wiki --skill maintain -a codex -y
Cursor
$ npx skills add Oshayr/LLM-Wiki --skill maintain -a cursor -y
Gemini CLI
$ npx skills add Oshayr/LLM-Wiki --skill maintain -a gemini-cli -y
GitHub Copilot
$ npx skills add Oshayr/LLM-Wiki --skill maintain -a github-copilot -y
OpenCode
$ npx skills add Oshayr/LLM-Wiki --skill maintain -a opencode -y
Windsurf
$ npx skills add Oshayr/LLM-Wiki --skill maintain -a windsurf -y
SKILL.md
maintain.SKILL.mdname: wiki-maintain
description: "Wiki maintenance — lint broken links, merge near-duplicates, upgrade confidence, flag stale pages via freshness tiers, gap analysis, concept synthesis. Use on: 'wiki maintenance', 'wiki cleanup', 'fix wiki', 'wiki health', 'check wiki', 'consolidate wiki'."
Wiki Maintain
Comprehensive wiki maintenance: lint, deduplicate, upgrade, and analyze.
Resolve `.wiki/` from plugin install scope. If not found, say "No wiki found."
Arguments
- **`/wiki-maintain`** — run full maintenance (all steps below)
- **`/wiki-maintain lint`** — only fix broken links, missing frontmatter, orphans
- **`/wiki-maintain dedup`** — only find and merge near-duplicate pages
- **`/wiki-maintain gaps`** — only analyze knowledge gaps and missing coverage
Full Maintenance Steps
1. Lint
Launch `wiki-auditor` agent to:
- Find `[[wiki-links]]` that point to non-existent pages — list for user
- Fix missing frontmatter fields (add defaults)
- Find orphan pages (no incoming links) — suggest connections
- Remove dead index entries
- Fix stale `updated:` dates on pages that were modified but not date-bumped
2. Deduplicate
Find pages with >60% slug token overlap (Jaccard similarity on slug words split by `-`):
- Report pairs for review
- For confirmed duplicates: auto-merge into one page, update all backlinks, add redirect note to absorbed page
3. Confidence Upgrade
Pages with 3+ independent sources in frontmatter get upgraded:
- `low` → `medium` (if 2+ sources)
- `medium` → `high` (if 3+ corroborating sources)
- Run `fact-checker` agent on high-confidence pages to verify claims against external sources
- Write the reason in log.md
4. Stale Detection (Freshness Tiers)
Pages are evaluated against an intelligent freshness system — not a flat threshold. Each page has a TTL based on its content type:
| Tier | TTL | Examples | |------|-----|----------| | `live` | 15 min | stock prices, live scores, server status, deployment state | | `breaking` | 1-6 hours | breaking news, incident updates, release announcements | | `current` | 1-3 days | news articles, current events, trending topics | | `fast` | 1-4 weeks | AI/LLM/MCP, API changes, model benchmarks | | `moderate` | 1-3 months | software versions, frameworks, libraries, tools | | `standard` | 6 months | general knowledge, how-to guides (default) | | `academic` | 1 year | research papers, studies, formal publications | | `evergreen` | 5 years | history, biographies, foundational concepts, laws, theorems | | `permanent` | never | personal notes, ideas, memories, journal entries |
Resolution order: 1. Explicit `freshness_tier:` in page frontmatter (user override) 2. Explicit `ttl:` in frontmatter (custom duration like `ttl: 30m` or `ttl: 2d`) 3. Auto-classification from tags, type, and content keywords
For stale pages:
- Add `stale: true` to frontmatter
- Suggest running `/wiki-write --refresh-stale` or `/wiki-read` to refresh
5. Concept Auto-Generation
Detect patterns spanning 3+ pages:
- Find groups of pages that share 3+ common `[[wiki-links]]` targets
- For each cluster: suggest a synthesis article that connects the concepts
- If user approves: generate the synthesis page via `wiki-writer` agent
6. Regenerate Index
Rebuild `.wiki/index.md` from all pages:
- Group by `type:` field (concept, entity, source, idea, status, etc.)
- Alphabetical within each group
- Include confidence badge and one-line description
7. Report
Print summary: broken links fixed, duplicates found, confidence upgrades, stale flags, concepts suggested, index updated.
Read more
name: wiki-maintain description: "Wiki maintenance — lint broken links, merge near-duplicates, upgrade confidence, flag stale pages via freshness tiers, gap analysis, concept synthesis. Use on: 'wiki maintenance', 'wiki cleanup', 'fix wiki', 'wiki health', 'check wiki', 'consolidate wiki'."
Wiki Maintain
Comprehensive wiki maintenance: lint, deduplicate, upgrade, and analyze.
Resolve `.wiki/` from plugin install scope. If not found, say "No wiki found."
Arguments
- **`/wiki-maintain`** — run full maintenance (all steps below)
- **`/wiki-maintain lint`** — only fix broken links, missing frontmatter, orphans
- **`/wiki-maintain dedup`** — only find and merge near-duplicate pages
- **`/wiki-maintain gaps`** — only analyze knowledge gaps and missing coverage
Full Maintenance Steps
1. Lint
Launch `wiki-auditor` agent to:
- Find `[[wiki-links]]` that point to non-existent pages — list for user
- Fix missing frontmatter fields (add defaults)
- Find orphan pages (no incoming links) — suggest connections
- Remove dead index entries
- Fix stale `updated:` dates on pages that were modified but not date-bumped
2. Deduplicate
Find pages with >60% slug token overlap (Jaccard similarity on slug words split by `-`):
- Report pairs for review
- For confirmed duplicates: auto-merge into one page, update all backlinks, add redirect note to absorbed page
3. Confidence Upgrade
Pages with 3+ independent sources in frontmatter get upgraded:
- `low` → `medium` (if 2+ sources)
- `medium` → `high` (if 3+ corroborating sources)
- Run `fact-checker` agent on high-confidence pages to verify claims against external sources
- Write the reason in log.md
4. Stale Detection (Freshness Tiers)
Pages are evaluated against an intelligent freshness system — not a flat threshold. Each page has a TTL based on its content type:
| Tier | TTL | Examples | |------|-----|----------| | `live` | 15 min | stock prices, live scores, server status, deployment state | | `breaking` | 1-6 hours | breaking news, incident updates, release announcements | | `current` | 1-3 days | news articles, current events, trending topics | | `fast` | 1-4 weeks | AI/LLM/MCP, API changes, model benchmarks | | `moderate` | 1-3 months | software versions, frameworks, libraries, tools | | `standard` | 6 months | general knowledge, how-to guides (default) | | `academic` | 1 year | research papers, studies, formal publications | | `evergreen` | 5 years | history, biographies, foundational concepts, laws, theorems | | `permanent` | never | personal notes, ideas, memories, journal entries |
Resolution order: 1. Explicit `freshness_tier:` in page frontmatter (user override) 2. Explicit `ttl:` in frontmatter (custom duration like `ttl: 30m` or `ttl: 2d`) 3. Auto-classification from tags, type, and content keywords
For stale pages:
- Add `stale: true` to frontmatter
- Suggest running `/wiki-write --refresh-stale` or `/wiki-read` to refresh
5. Concept Auto-Generation
Detect patterns spanning 3+ pages:
- Find groups of pages that share 3+ common `[[wiki-links]]` targets
- For each cluster: suggest a synthesis article that connects the concepts
- If user approves: generate the synthesis page via `wiki-writer` agent
6. Regenerate Index
Rebuild `.wiki/index.md` from all pages:
- Group by `type:` field (concept, entity, source, idea, status, etc.)
- Alphabetical within each group
- Include confidence badge and one-line description
7. Report
Print summary: broken links fixed, duplicates found, confidence upgrades, stale flags, concepts suggested, index updated.
An autonomous knowledge base that grows as you work. LLM Wiki is a Claude Code plugin that captures research, ideas, and decisions into an interlinked wiki with semantic search, automatic research, and a Wikipedia-style web UI.
More skills from other plugins.
codebase-wiki
How to work in a Codebase Wiki project (the `codebase-wiki` starter pack): an agent-authored, source-grounded wiki of the…
academic-paper-reviewe…
Simulates academic peer review, evaluating papers across Originality, Methodology, Results, and Writing to provide Major/Minor…
Questions about maintain.
- Which coding agents can use maintain?
- maintain installs in Claude Code, Codex, Cursor, Gemini CLI, GitHub Copilot, OpenCode and Windsurf with the skills CLI, using each agent's flag shown on this page. Flowy's auto-invocation, which fires a skill when your prompt matches, runs in Claude Code only.

