What does this skill do, and when should you use it?
wiki-lint is the maintenance skill in claude-obsidian for checking an Obsidian vault built from plain Markdown. It scans for orphan pages, dead wikilinks, stale claims, missing cross-references, frontmatter gaps, empty sections, and stale index entries, then writes a dated lint report. It can also create or update Dataview dashboards and a Canvas domain map. When DragonScale is enabled, it adds address validation and semantic duplicate detection; automatic fixes require confirmation first.
Reads .vault-meta/transport.json to select CLI, MCP, or filesystem transport, then checks orphan pages, dead links, stale claims, missing pages, missing cross-references, frontmatter gaps, empty sections, stale index entries, naming conventions, and writing-style issues. It writes wiki/meta/lint-report-YYYY-MM-DD.md and creates or updates wiki/meta/dashboard.md and wiki/meta/overview.canvas. With DragonScale enabled, it can run allocate-address.sh --peek and tiling-check.py to validate address formats, uniqueness, counter consistency, address mappings, and semantically similar pages.
- A researcher with an Obsidian vault who wants weekly detection of orphan pages and broken wikilinks.
- A knowledge worker who regularly ingests articles, books, or other sources and needs to find missing cross-references.
- A Markdown-based personal knowledge base with extensive frontmatter that needs checks for missing fields and empty sections.
- A DragonScale user managing a large vault who needs address-collision, counter-drift, or duplicate-page checks.
How do you install this skill?
- The skill says it writes a single lint report but also requires creating or updating dashboard.md and overview.canvas; confirm this additional write scope and preserve recoverable backups.
- Semantic tiling processes page bodies; even with the local-Ollama default, confirm local service/model availability and the vault's sensitive-data policy.
- Do not assume completion when Dataview, Obsidian CLI, MCP, Python, Ollama, or referenced scripts are missing; report each skipped, failed, or unverified check explicitly.
- No dedicated wiki-lint execution evidence is provided. Treat findings as recommendations requiring user review, especially stale-claim, missing-page, and duplicate-page judgments.
- Shell / CLI
- Network access
- Local filesystem
- MCP Server
obsidian-cliPython 3ollama
The repository bundles 15 skills, and the source does not document a standalone wiki-lint installation. To install the public collection, run git clone https://github.com/AgriciDaniel/claude-obsidian, then cd claude-obsidian and bash bin/setup-vault.sh. Alternatively, add the marketplace with claude plugin marketplace add AgriciDaniel/claude-obsidian, then install it with claude plugin install claude-obsidian@agricidaniel-claude-obsidian.
tmp="$(mktemp -d)"
git clone --depth 1 https://github.com/AgriciDaniel/claude-obsidian.git "$tmp"
mkdir -p ~/.claude/skills
cp -R "$tmp/skills/wiki-lint" ~/.claude/skills/
rm -rf "$tmp"Generated from the source repository and skill path; it copies only this skill's folder. If the author's install steps above differ, follow those first. To scope it to one project, replace ~/.claude/skills with that project's .claude/skills.
How do you use this skill?
Once installed, send your agent any of these to trigger it:
- Should I fix these automatically, or do you want to review each one?
In a configured vault, use a trigger such as lint the wiki, health check, clean up wiki, check the wiki, wiki maintenance, find orphans, or wiki audit. The skill recommends running after every 10–15 ingests or weekly, with output at wiki/meta/lint-report-YYYY-MM-DD.md. After reviewing the report, answer: “Should I fix these automatically, or do you want to review each one?” The source does not document additional standalone invocation steps for wiki-lint.
What are this skill's strengths and limitations?
- Covers multiple maintenance categories, including orphans, dead links, stale claims, metadata gaps, and empty sections.
- Defines a dated Markdown report path and format for persistent vault tracking.
- Supports CLI, MCP, and filesystem transport paths.
- DragonScale adds address-consistency and semantic-duplicate checks when enabled.
- It primarily diagnoses and reports problems rather than resolving every issue automatically.
- Deleting orphan pages, resolving contradictions, and merging duplicates require human review.
- Semantic tiling depends on Python, ollama, and a local embedding model, and hard-fails above 5,000 pages.
- The source provides no platform verification evidence beyond the Claude Code and Codex compatibility baseline.
How does this skill compare with similar options?
Side by side with related skills; every score comes from the same FSRS standard.
| Skill | FS score | Stars | Last updated | License |
|---|---|---|---|---|
| Obsidian Wiki Health Check this page | 48 · Use with care | ★ 16k | 1mo ago | MIT |
| Obsidian Markdown Editor | 57 · Use with care | ★ 50k | 1mo ago | MIT |
| Obsidian Vault Organizer | 49 · Use with care | ★ 10k | 13d ago | MIT |
| Wiki Log Folder | 55 · Use with care | ★ 16k | 1mo ago | MIT |
| Autonomous Research Filing | 52 · Use with care | ★ 16k | 1mo ago | MIT |
The README compares the overall claude-obsidian project with Smart Connections and Copilot; wiki-lint specifically corresponds to the project's vault-maintenance capability, not the full feature set of those products. It also works alongside Dataview: the skill can update a Dataview dashboard, while the repository presents Bases as the primary dashboard option.
How did FollowSkills review this skill?
The skill explicitly reads before writing, shows the report first, asks before auto-fixing, and defaults semantic tiling to local Ollama while requiring an explicit override for remote access and disclosing that page bodies are POSTed; however, the claim of writing a single report conflicts with later unconditional dashboard/canvas creation or updates, while auto-fix boundaries, rollback, dependency trust, and sensitive-data handling remain incomplete, so points are deducted.
The check order, report format, exit codes, and abnormal branches are fairly specific, with indirect support from repository CI and related script tests; however, the selected skill has no dedicated tests, and Dataview, Obsidian CLI, MCP, Ollama, and referenced paths may be unavailable. Unified diagnostics and reproducibility for dependency failures are thin, so the static score stays below 10.
Triggers, primary scenarios, and scan scope are clear, with filesystem, CLI, MCP, and local-Ollama paths described; nevertheless, input prerequisites, output boundaries, non-fit cases, and compatibility across Obsidian and transport configurations are under-specified. Chinese-content support has only indirect test evidence, and mainland-China reachability is not established, warranting deductions.
The skill has frontmatter, layered sections, report templates, configuration examples, exit-code documentation, security assumptions, and repository MIT licensing; but versioning/changelog, maintenance ownership, and update paths rely mainly on the README, installation and dependency notes are distributed across other files, and the dashboard/canvas write behavior conflicts with the transport description, so full marks are not justified.
Eight baseline checks, optional address validation, semantic tiling, report templates, and confirmation before fixes cover the core wiki-maintenance objective, and the expected report is directly usable; however, static files do not prove the skill's key paths execute successfully. Finding quality for stale claims, missing pages, and Dataview/canvas outputs still requires review, so the static ceiling supports 7.
Repository CI, related script tests, explicit report structures, and enumerated error codes provide limited auditability; however, there are no wiki-lint-specific tests or real execution results for the selected skill, and README benchmark/performance claims cannot substitute for reproducibility of this skill, so the score remains below the static maximum.
Open a dimension to read why it scored that way
Evidence confidence：Low — Mostly static review, author material or a limited demo; useful for discovery, not high-risk decisions.
See the full review method →