llm-wiki

Karpathy's LLM Wiki: build/query interlinked markdown KB.

Quality

50%

Does it follow best practices?

Run evals on this skill

Adds up to 20 points to the overall score

View guide

Securityby

High

Do not use without reviewing

Fix and improve this skill with Tessl

tessl review fix ./skills/research/llm-wiki/SKILL.md

Quality

Content

50%

Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.

This is a well-structured, genuinely actionable skill with strong numbered workflows and built-in validation (orientation-first, sha256 drift detection, lint), but it is monolithic and verbose with inline templates/headless setup, includes a pseudocode lint example, and lacks verification checkpoints on its destructive and batch operations.

Suggestions

Move the full SCHEMA.md/index.md/log.md templates and the Obsidian-headless/systemd setup into reference files (e.g., references/templates.md, references/obsidian-sync.md) and link to them one level deep to improve progressive disclosure and conciseness.

Replace the pseudocode orphan-scan block with executable code (or explicitly justify the pseudocode) so the lint guidance is copy-paste ready.

Add explicit verification checkpoints to destructive and batch workflows: e.g., after archiving, re-run a link check to confirm no broken wikilinks and the index is consistent; after bulk ingest, run a quick lint/duplicate check before reporting completion.

Dimension	Reasoning	Score
Conciseness	The ~500-line body is mostly domain-specific procedure that earns its place, but the inline SCHEMA/index/log templates and the full Obsidian-headless + systemd setup are verbose and could be tightened or split out; it fits 'mostly efficient but includes some unnecessary explanation or could be tightened' rather than the lean level 3.	2 / 3
Actionability	Concrete frontmatter schemas, naming rules, thresholds, and bash search commands are largely copy-paste ready, but the lint orphan-scan code block is pseudocode (comments only, no real implementation) and several snippets use tool-call placeholders like 'offset=<last 30 lines>', matching 'some concrete guidance but incomplete; pseudocode instead of executable code'.	2 / 3
Workflow Clarity	Ingest/query/lint/init use clear numbered sequences with strong orientation-first validation and a lint feedback loop, but the destructive archiving workflow and batch bulk-ingest lack explicit verification checkpoints (e.g., re-lint after archive, verify no duplicates after bulk ingest), so per the rubric workflow_clarity is capped at 2.	2 / 3
Progressive Disclosure	Sections are well organized, but the skill is a monolithic single file with no reference bundle; the full SCHEMA/index/log templates and the Obsidian-headless setup are inline content that should be separate, fitting 'some structure but content that should be separate is inline' rather than the one-level-deep reference structure of level 3.	2 / 3
	Total	8 / 12 Passed

Description

50%

Based on the skill's description, can an agent find and select it at the right time? Clear, specific descriptions lead to better discovery.

The description is concise and identifies a real niche with a couple of concrete actions, but it omits any explicit 'Use when...' trigger guidance and offers only modest trigger-term coverage, leaving all dimensions at the mid-level.

Suggestions

Add an explicit trigger clause, e.g. 'Use when the user asks to build or maintain a wiki/knowledge base, ingest sources, query an existing wiki, or lint/audit their notes.'

Broaden natural trigger terms to include 'notes' and 'research' alongside 'wiki' and 'knowledge base' to improve coverage and reduce overlap with the obsidian skill.

Add one more concrete action (e.g., 'cross-reference', 'lint', or 'audit') to lift specificity toward listing multiple specific actions.

Dimension	Reasoning	Score
Specificity	Names the domain ('interlinked markdown KB') and two concrete actions ('build/query'), matching the 'names domain and some actions, but not comprehensive' anchor; it does not list multiple distinct concrete actions required for a 3, yet is far from the vague 'helps with documents' level 1.	2 / 3
Completeness	It states what the skill does ('build/query interlinked markdown KB') but has no 'Use when...' clause or explicit trigger guidance, so per the judging guidelines completeness is capped at 2 ('has what, but when is missing or only implied').	2 / 3
Trigger Term Quality	Natural terms 'wiki', 'knowledge base'/'KB', and 'markdown' are present, but coverage is thin and misses common variations users would say like 'notes' or 'research', fitting 'some relevant keywords but missing common variations'.	2 / 3
Distinctiveness Conflict Risk	The named 'Karpathy's LLM Wiki' pattern gives it a niche, but 'wiki'/'knowledge base'/'markdown KB' overlaps with obsidian and notes-style skills with no disambiguating trigger, fitting 'somewhat specific but could still overlap with similar skills'.	2 / 3
	Total	8 / 12 Passed

Validation

75%

Warnings & errors only

Checks the skill against the spec for correct structure and formatting. All validation checks must pass before discovery and implementation can be scored.

Validation — 12 / 16 Passed

Validation for skill structure

Criteria	Description	Result
skill_md_line_count	SKILL.md is long (508 lines); consider splitting into references/ and linking	Warning
metadata_version	'metadata.version' is missing	Warning
metadata_field	'metadata' should map string keys to string values	Warning
frontmatter_unknown_keys	Unknown frontmatter key(s) found; consider removing or moving to metadata	Warning

	Total	12 / 16 Passed

Repository: NousResearch/hermes-agent
Path: skills/research/llm-wiki/SKILL.md
Commit: d0d116b

Reviewed: about 14 hours ago

Table of Contents

Discovery Implementation Validation

Is this your skill?

If you maintain this skill, you can claim it as your own. Once claimed, you can manage eval scenarios, bundle related skills, attach documentation or rules, and ensure cross-agent compatibility.