Agent skill

Kb

by compozy in compozy/kb

Operate the kb CLI to build and maintain topic-based Karpathy-style knowledge bases: gated ingestion, decision-model classification, linking, retrieval and review, wiki compilation, lint-and-heal…

MITAuto-check passedKnowledge Management

Install Kb

skills CLI
$ npx skills add compozy/kb --skill kb -a claude-code

Project install by default; add -g for ~/.claude/skills/.

GitHub CLI
$ gh skill install compozy/kb kb --agent claude-code

Project scope by default; add --scope user for a personal install. Needs GitHub CLI 2.90.0 or later (public preview).

Manual copy
$ git clone --depth 1 https://github.com/compozy/kb.git skills-src && mkdir -p .claude/skills && cp -r skills-src/skills/kb .claude/skills/kb && rm -rf skills-src

Use ~/.claude/skills/ instead of .claude/skills for a personal install. The folder must contain SKILL.md.

Claude Code skills documentation · loads skills from .claude/skills/

Facts

Skill name
kb
GitHub stars
110
Token cost
~4.2k tokens
SKILL.md length
2,078 words
Files
16 (incl. references)
Skills in repo
1
Repo updated
First seen
Licence
MIT

At a glance

Operate the kb CLI to build and maintain topic-based Karpathy-style knowledge bases: gated ingestion, decision-model classification, linking, retrieval and review, wiki compilation, lint-and-heal…

  • Works in 8 steps: Read references/compilation-guide.md for… → Find candidate sources: kb find ""… → Load the candidate sources fully, and… → …
  • Kb commands and KB vault workflows
  • SKILL.md covers Prerequisites, Command map, Working with the decision model and KB maintenance procedures, plus 3 more sections
  • Calls npm; needs OPENROUTER_API_KEY and FIRECRAWL_API_KEY

What it does

Kb is an agent skill from compozy/kb. Operate the kb CLI to build and maintain topic-based Karpathy-style knowledge bases: gated ingestion, decision-model classification, linking, retrieval and review, wiki compilation, lint-and-heal, OKF bundles, QMD search, and codebase inspection. Use for kb commands and KB vault workflows; not for general code review, Go builds, or application code.

Its SKILL.md is about 4.2k tokens, which your agent loads only when the skill is triggered. The skill folder holds 16 other files, including reference files (for example `references/architecture.md`, `references/cleaning-a-topic.md` and `references/cli-ingest-codebase.md`).

It sits in Knowledge Management, covering Knowledge bases, Linting and formatting and LLM wikis. The licence is MIT.

When your agent uses it

  • Kb commands and KB vault workflows
  • Not for general code review
  • Application code

Example prompts

  • “/kb”

Requirements

  • Node.js
  • A credential in OPENROUTER_API_KEY
  • A credential in FIRECRAWL_API_KEY

Workflow steps

8 steps, taken from the first numbered list in SKILL.md.

  1. Read references/compilation-guide.md for length, style, wikilink density, sourcing and update rules.
  2. Find candidate sources: kb find "" --explain, sources whose concepts: or affects: name the article, kb search "" --topic , or…
  3. Load the candidate sources fully, and /wiki/index/Concept Index.md for existing articles and wikilink targets.
  4. Before drafting, present 3–5 key takeaways, the concepts the article introduces or updates, and any contradiction with existing articles…
  5. Write /wiki/concepts/.md (Obsidian Flavored Markdown; frontmatter from frontmatter-schemas.md; 3000–4000 words; Sources section). Updates…
  6. Run kb classify --only-missing (criterion and aliases for the new article), then kb link , and work kb review --queue link. kb finds the…
  7. Update the topic indexes (Procedure 2) and the current-articles list in /CLAUDE.md; re-index with kb index --topic when the topic uses QMD.
  8. Append ## [YYYY-MM-DD] compile | ( words, sources) to /log.md.

What it can do on your machine

Read from SKILL.md and the folder at commit d7c8261. It shows what the files ask for, not the result of running them.

  • Tool permissions

    Pre-approves nothing: there is no allowed-tools line, so your agent's usual permission prompts apply.

    From allowed-tools in the SKILL.md frontmatter.

  • Runs code

    Shell commands in SKILL.md call:

    • npm

    From the folder's file list and the shell code blocks in SKILL.md.

  • Network

    Links to these hosts (documentation or services it may open):

    • github.com

    From URLs in SKILL.md, links to its own repository left out.

  • Credentials

    Names these keys or tokens, usually read from environment variables:

    • OPENROUTER_API_KEY
    • FIRECRAWL_API_KEY

    From names ending in _API_KEY, _TOKEN, _SECRET, _KEY or _PASSWORD in SKILL.md.

Context cost

Kb loads about 4.2k tokens when it runs, and up to ~36k if it reads all its reference files. Until then it costs about 89 tokens; SKILL.md has 2,078 words of instructions outside code blocks.

Always · name and description, kept in context so the agent knows when to use it
~89
When it runs · the whole SKILL.md, loaded when a task matches
~4.2k
With references · SKILL.md plus every file in references/, read only if the agent opens them
~36k

Estimates: characters ÷ 4, the usual rule of thumb; real counts depend on the model's tokenizer. Scripts and assets cost tokens only if the agent reads them.

Safety

Auto-check passed

The automated check found no risky patterns in SKILL.md.

Automated static check — not a guarantee. Review scripts before installing. It scans the text of SKILL.md for risky patterns (piping downloads into a shell, reading credential files, hidden Unicode, destructive commands); files beside SKILL.md are not scanned.

SKILL.md

The full file from compozy/kb at commit d7c8261, republished under its MIT licence (© compozy). 2,078 words, ~4,227 tokens.

Download SKILL.mdSave it as .claude/skills/kb/SKILL.md (or your agent's skills folder). This skill also uses 15 other files; get the full folder from GitHub.
name
kb
description
Operate the `kb` CLI to build and maintain topic-based Karpathy-style knowledge bases: gated ingestion, decision-model classification, linking, retrieval and review, wiki compilation, lint-and-heal, OKF bundles, QMD search, and codebase inspection. Use for kb commands and KB vault workflows; not for general code review, Go builds, or application code.

kb CLI and Knowledge Base Pattern

kb maintains an Obsidian markdown knowledge base in the Karpathy LLM Wiki pattern. kb ingests sources, and its decision model gates, classifies and links them. You, the agent, compile wiki articles from the sources, answer questions from the wiki and file the answers back, and run lint-and-heal passes.

  • Topic. Each topic is a folder in the vault: at the root (go-best-practices/) or nested by [vault].topic_globs (harness/goclaw/). CLAUDE.md is the topic marker and schema document (scope, conventions, current articles, gaps; AGENTS.md may symlink to it). topic.yaml holds slug, title, domain, mode and the selection contract. kb renders the contract into CLAUDE.md under ## Selection contract; change it with kb topic contract, never in CLAUDE.md.
  • Mode. mode: wiki (default; raw/, wiki/, outputs/, bases/, log.md, the ingest → compile → query → lint loop) or mode: okf (a flat, portable Open Knowledge Format bundle of typed concepts, filled by kb promote).
  • Division of labour. The decision model only judges text (yes/no, choice, score) and a small generation model only writes short literals (summaries, criteria, aliases, drafts); code applies thresholds and writes files. You write and update article prose, decide what to ingest, draft or edit contracts, and give the human verdicts in kb review.

references/architecture.md has the rationale (compiler analogy, context window vs RAG, multi-topic vaults).

Prerequisites

  • kb version works. qmd --version for kb index/kb search and optional vector candidates (npm install -g @tobilu/qmd).
  • OPENROUTER_API_KEY for every command that ingests (except codebase), classifies, links, finds, drafts or accepts a contract, builds a vocabulary, accepts or rejects review items, or suggests an OKF type. Those commands refuse to start without it. lint, inspect, index, search, topic new|list|info, topic contract --import-claude, review list|import-links|import-labels|calibrate, find --facets, ingest codebase and migrate never call a model.
  • FIRECRAWL_API_KEY for kb ingest url; yt-dlp (and an STT key for --transcribe auto|stt) for YouTube and Instagram.
  • A repo-root vault has kb.toml at its root:
    toml
    [vault]
    root = "."
    topic_globs = ["*", "harness/*"]

Command map

IntentCommandDetails
New wiki topic / OKF bundlekb topic new <slug> <title> <domain> [--mode okf]okf-mode.md
List topics, topic metadatakb topic list, kb topic info <topic>
Ingest a URL, URL list, file, YouTube video, channel, Instagram reel, bookmark clusterkb ingest url|file|youtube|channel|instagram|bookmarks <input> --topic <topic>cli-ingest-sources.md
Write, import or accept the selection contractkb topic contract <topic> --draft|--import-claude|--acceptselection-contract.md
Bootstrap concepts for a topic without articleskb topic vocabulary <topic> --draft, then --acceptdecision-workflow.md
Classify documents (facets, summaries, criteria, aliases)kb classify <topic> [--only-missing]decision-workflow.md
Write typed relations and backlinkskb link <topic> [--dry-run]decision-workflow.md
Find the documents that answer a questionkb find <topic> "<question>" --explaindecision-workflow.md
Count documents per genre, relevance, depth, conceptkb find --facets <topic>
Work the review queueskb review <topic> [--queue q]; kb review accept|reject --topic <topic> <id>...decision-workflow.md
Import screening files or human links as labelskb review import-labels <topic> --from <file> ...; kb review import-links <topic>decision-workflow.md
Calibrate thresholdskb review calibrate <topic> [--write]decision-workflow.md
Clean a topic that predates the decision modelfive stepscleaning-a-topic.md
Lint a topickb lint <topic> [--save]lint-procedure.md
Index and search with QMDkb index --topic <topic>; kb search "<query>" --topic <topic> [--lex|--vec] --format jsoncli-search-index.md
Distill a wiki concept into an OKF bundlekb promote <wiki-doc> --to <okf-topic> [--type <Type>]okf-mode.md
Check an OKF bundlekb okf check <okf-topic> [--strict] [--decide]okf-mode.md
Analyze a codebase, then inspect itkb ingest codebase <path> --topic <topic> --progress never; kb inspect <subcommand> --format jsoncli-ingest-codebase.md, cli-inspect.md
Move legacy transcriptskb migrate transcripts --topic <topic>cli-ingest-sources.md

Use path-relative topic ids for nested topics (--topic harness/goclaw). For parsing, use --format json where a command offers it and read stdout only; stderr carries diagnostics and run summaries (output-formats.md).

Working with the decision model

These rules apply to every decision-backed command; references/decision-workflow.md has flags, thresholds, queue actions, calibration and the .decisions/ files.

  • Budget. --budget <usd> caps a run (default US$ 1). --budget 0 uses cached answers only. Past the budget, items are undecided:budget: re-run to continue from the cache in <topic>/.decisions/receipts.jsonl.
  • Failures are never a "no". undecided:<reason> and not_checked:<reason> leave a document as it was, and an undecided ingest judgment writes triage: review with triage_reason: undecided. Report these statuses from the run summary; re-run instead of concluding anything from them.
  • Shadow before apply. Relevance gates and body-link insertion run in shadow (recorded as review items, nothing moved) until the topic has an accepted contract and a current calibration or decisions.gates: apply / decisions.mode: apply. Quality gates and exact dedupe apply from the start. Never switch a topic to apply without the user.
  • Nothing is deleted. Gated sources move to raw/_quarantine/ with a ledger of every index line and frontmatter entry removed, so kb review accept restores them byte for byte. Quarantine and restore only through kb review, never by moving files.
  • Frontmatter ownership. kb writes its own keys (triage, genre, depth, relevance, concepts, summary, criterion, entities, questions, quality, relation lists, aliases, ...; frontmatter-schemas.md). A key you or the user edit becomes user-owned and kb stops writing it (lint: key-conflict); relation lists and aliases only get appended to. locked: true makes kb skip a file entirely. topic.yaml decisions.exclude globs keep files out of every model call.
  • Human verdicts. Review items are yours or the user's to decide. Read items before bulk-accepting (--above, --topic-wide); several wrong remove items mean the contract is too narrow. Every verdict becomes a calibration label.

KB maintenance procedures

Procedure 1: Compile a wiki article
  1. Read references/compilation-guide.md for length, style, wikilink density, sourcing and update rules.
  2. Find candidate sources: kb find <topic> "<what the article must cover>" --explain, sources whose concepts: or affects: name the article, kb search "<phrase>" --topic <topic>, or <topic>/wiki/index/Source Index.md. For a stub (stage: stub), its criterion says what the article must cover.
  3. Load the candidate sources fully, and <topic>/wiki/index/Concept Index.md for existing articles and wikilink targets.
  4. Before drafting, present 3–5 key takeaways, the concepts the article introduces or updates, and any contradiction with existing articles; ask what to emphasize, unless the user asked for autonomous compilation.
  5. Write <topic>/wiki/concepts/<Article Title>.md (Obsidian Flavored Markdown; frontmatter from frontmatter-schemas.md; 3000–4000 words; Sources section). Updates to an existing article use the Current / Proposed / Reason / Source diff format from compilation-guide.md.
  6. Run kb classify <topic> --only-missing (criterion and aliases for the new article), then kb link <topic>, and work kb review <topic> --queue link. kb finds the mentions and writes the backlinks; you do not hand-audit them.
  7. Update the topic indexes (Procedure 2) and the current-articles list in <topic>/CLAUDE.md; re-index with kb index --topic <topic> when the topic uses QMD.
  8. Append ## [YYYY-MM-DD] compile | <Article Title> (<words> words, <N> sources) to <topic>/log.md.
Procedure 2: Maintain topic indexes

After adding, renaming or removing a wiki article, update <topic>/wiki/index/Dashboard.md (counts, featured sections, Base embeds), Concept Index.md (alphabetical row with a one-line summary) and Source Index.md (every cited source with a wikilink back to the article).

Procedure 3: Query the wiki and file back the answer
  1. Read the topic's Concept Index.md first, then locate articles: the index at small scale (< 30 articles), otherwise kb find <topic> "<question>" --explain, plus kb search or grep for exact terms. Answer from the wiki, never from general knowledge; a contradiction between the two is itself a finding.
  2. Read the identified articles in full, following one level of [[wikilinks]] when relevant. Pull a raw source only to verify an ambiguous claim.
  3. Answer with a [[wikilink]] citation for every factual claim, note agreements and disagreements between articles, and state gaps explicitly ("the wiki has no article on X"). Match the format to the question: prose, comparison table, numbered steps, Known / Open questions / Gaps, or a diagram (tooling-tips.md for Marp).
  4. Save the answer to <topic>/outputs/queries/<YYYY-MM-DD> <Question Slug>.md (type: output, stage: query, informed_by: wikilinks) and call out insights to absorb on the next compile.
  5. When the answer contradicts or extends an article, recompile it (Procedure 1). When the synthesis is durable (a comparison, a trade-off analysis, a new concept), promote it to wiki/concepts/ with Procedure 1 standards.
  6. Append ## [YYYY-MM-DD] query | <Question Slug> (and ## [YYYY-MM-DD] promote | <Title> when promoted) to log.md.
Show full SKILL.md (742 more words)Show less
Procedure 4: Lint and heal

Run kb lint <topic> --save (never calls a model; the report goes to <topic>/outputs/reports/). Propose each fix as a diff before applying; do not batch-apply. references/lint-procedure.md maps every issue kind to its fix (dead links, orphans, missing sources, needs-compile, link-to-quarantined, key-conflict, the review-queue counts, missing contract, vocabulary or criterion, unclassified, format) and covers the semantic checks lint cannot do. Append ## [YYYY-MM-DD] lint | <N> issues found, <M> fixed to log.md.

Procedure 5: Log entries

kb appends log.md entries for ingest (one per run, with gate counts and cost), lint --save and kb promote (in the OKF bundle's log). Add the others yourself as one H2 line each, newest last, so the log stays greppable:

markdown
## [YYYY-MM-DD] <compile|query|promote|split> | <short description>

promote here is the wiki-internal query → wiki promotion, not kb promote. grep "^## \[" <topic>/log.md | tail -10 shows recent activity.

Procedure 6: Distill a wiki concept into an OKF bundle
  1. Target an existing mode: okf topic (or create one with kb topic new ... --mode okf); kb promote --to must be an OKF bundle.
  2. Promote a compiled <topic>/wiki/concepts/<Article>.md: kb promote "<path>" --to <okf-topic> [--type "<Type>"] [--description "<one line>"]. Without --type, the decision model suggests one from [okf].types; a weak suggestion lists candidates and asks for --type.
  3. Promote the targets listed in unresolvedLinks when those links should resolve.
  4. Run kb okf check <okf-topic> (--strict in CI). No manual log entry is needed.

kb promote is mechanical: it remaps frontmatter, rewrites [[wikilinks]] to relative links, regenerates index.md and leaves the source untouched; condensing prose is your job. references/okf-mode.md has the producer-field contract and conformance rules.

Procedure 7: Set up the decision workflow on a topic
  • New topic, after the first ingests: kb topic contract <topic> --draft (or hand-write contract_draft following selection-contract.md), edit it, then kb topic contract <topic> --accept. Show the impact preview (bands, per-folder counts, the 20 most off-topic documents) to the user before the slug is typed; only the user decides whether the contract describes the collection. From then on every ingest is gated, classified and linked; check kb review <topic> after bulk ingests.
  • Topic that predates the decision model: follow references/cleaning-a-topic.md (contract, labels, classify and vocabulary, the recapture then remove queues, link).
  • Calibration: once the topic has ≥ 30 relevance labels given in kb review on the dev split and ≥ 10 on holdout, run kb review calibrate <topic>, show dev and holdout precision, and --write only with the user's approval. A later contract change makes the calibration stale and relevance gates return to shadow until it is redone.

Codebase analysis

Run kb ingest codebase <path> --topic <topic> --progress never before any kb inspect command on that topic; parse its stdout JSON (topicSlug, vaultPath, topicPath, diagnostics) and treat stderr stage logs as diagnostics, not failures. Codebase snapshots never enter the decision pipeline. Inspect subcommands (smells, dead-code, complexity, blast-radius, coupling, circular-deps, symbol, file, backlinks, deps) take --format json, --vault and --topic; inspect file needs the exact source-relative path (src/config.ts). Flags and schemas: cli-ingest-codebase.md, cli-inspect.md.

Constraints

  • Change the selection contract only through kb topic contract or topic.yaml contract_draft; the CLAUDE.md section is overwritten on accept.
  • Do not hand-edit kb-owned frontmatter keys unless you mean to take them over, and never edit files under <topic>/.decisions/.
  • Do not move, delete or restore gated sources by hand; use kb review accept|reject.
  • Do not set decisions.gates: apply or decisions.mode: apply without an impact preview, a calibration report and the user's approval.
  • Hand-edit neither an OKF bundle's index.md (kb regenerates it) nor links inside OKF concepts into [[wikilinks]] or absolute paths; OKF uses relative markdown links.
  • kb search takes at most one of --lex / --vec; kb index rejects --force-embed with --embed=false.

References

ReadWhen
references/decision-workflow.mdFlags of the decision commands, gates, modes and thresholds, review-queue actions, calibration, topic extra questions, .decisions/ files
references/selection-contract.mdDrafting, importing or editing a contract
references/cleaning-a-topic.mdBringing an existing topic under the decision workflow
references/compilation-guide.mdWriting or updating a wiki article
references/frontmatter-schemas.mdFrontmatter of any document type, topic.yaml, kb-owned keys and ownership rules
references/lint-procedure.mdHealing lint issues and the semantic checks
references/cli-ingest-sources.mdSource-specific ingest flags and configuration (URLs, files, YouTube, channels, Instagram, bookmarks)
references/okf-mode.mdOKF bundles, kb promote, kb okf check
references/cli-search-index.mdkb index and kb search
references/cli-ingest-codebase.md, references/cli-inspect.mdCodebase ingestion and inspection
references/output-formats.mdOutput shapes and empty results
references/error-handling.mdAny error message or run-summary status
references/architecture.mdWhy the pattern works, scale, multi-topic vaults
references/tooling-tips.mdObsidian plugins, Web Clipper, Marp, Dataview

Companion skills for the LLM-driven phases: obsidian-markdown (article syntax), obsidian-bases (.base dashboards), obsidian-cli (the running vault).

© compozy, MIT. Rendered from Markdown: HTML in the file is shown as text, images as links, and headings moved down two levels. Raw file

Files

SKILL.md and 15 other files (references) in skills/kb of compozy/kb.

  • SKILL.md
  • references/architecture.md
  • references/cleaning-a-topic.md
  • references/cli-ingest-codebase.md
  • references/cli-ingest-sources.md
  • references/cli-inspect.md
  • references/cli-search-index.md
  • references/compilation-guide.md
  • references/decision-workflow.md
  • references/error-handling.md
  • references/frontmatter-schemas.md
  • references/lint-procedure.md
  • references/okf-mode.md
  • references/output-formats.md
  • references/selection-contract.md
  • references/tooling-tips.md

Open the folder on GitHubat commit d7c8261

Compare with similar skills

Kb next to the 5 skills that share the most tags, products or categories with it. Stars are the repository's; “used in” counts other GitHub owners with a copy.

Kb compared with similar skills
SkillStarsUsed inTokensAuto-checkLicenceRepo updated
Kb this skillcompozy/kb110—~4.2kAutomated safety check: PassMIT
LLM Wikilewislulu/llm-wiki-skill655—~3.7kAutomated safety check: PassNone
Knowledge Base Health Lintjason-effi-lab/karpathy-llm-wiki-vault715—~336Automated safety check: PassNone
LLM Wiki Knowledge GraphEgonex-AI/Understand-Anything85k1 repos~1.5kAutomated safety check: PassMIT
Wiki Builderrohitg00/pro-workflow2.9k—~1kAutomated safety check: PassNone
Arkon Editnduckmink/arkon1.5k—~1.6kAutomated safety check: PassCustom licence

Similar skills

  • LLM Wiki

    lewislulu/llm-wiki-skill

    Build and maintain a Karpathy-style LLM knowledge base — a self-compiling Obsidian markdown wiki where an Agent ingests raw sources, compiles cross-linked concept/entity/summary pages, answers…

    655 GitHub stars~3.7k tokensUpdated 5 mo ago
    Knowledge ManagementAuto-check passed
  • Knowledge Base Health Lint

    jason-effi-lab/karpathy-llm-wiki-vault

    Scans an Obsidian-style wiki for broken double-link references, orphan pages, files missing from the index and unresolved knowledge conflicts, then reports the findings.

    715 GitHub stars~336 tokensUpdated 5 mo ago
    Knowledge ManagementAuto-check passed
  • LLM Wiki Knowledge Graph

    Egonex-AI/Understand-Anything

    Detects a Karpathy-pattern LLM wiki and builds an interactive knowledge graph with entities, implicit relationships and topic clusters.

    85k GitHub starsUsed in 1 repo~1.5k tokens
    Knowledge ManagementAuto-check passed
  • Wiki Builder

    rohitg00/pro-workflow

    Start, structure, and grow a persistent research wiki indexed in pro-workflow's SQLite knowledge base.

    2.9k GitHub stars~1k tokensUpdated 8 days ago
    Knowledge ManagementAuto-check passed
  • Arkon Edit

    nduckmink/arkon

    Propose or directly apply edits to Arkon wiki pages, including proposing brand new pages.

    1.5k GitHub stars~1.6k tokensUpdated 4 mo ago
    Knowledge ManagementAuto-check passed
  • Research Wiki Builder

    dair-ai/dair-academy-plugins

    Creates and maintains configurable research wikis: scaffold a folder, add sources, compile pages and indexes, and file query answers back.

    614 GitHub stars~1.3k tokensUpdated 2 mo ago
    Knowledge ManagementAuto-check passed

Questions about Kb

What does Kb do?

Operate the kb CLI to build and maintain topic-based Karpathy-style knowledge bases: gated ingestion, decision-model classification, linking, retrieval and review, wiki compilation, lint-and-heal…. Kb is an agent skill from compozy/kb. Operate the kb CLI to build and maintain topic-based Karpathy-style knowledge bases: gated ingestion, decision-model classification, linking, retrieval and review, wiki compilation, lint-and-heal, OKF bundles, QMD search, and codebase inspection.

When should I use Kb?

Kb fits situations like: kb commands and KB vault workflows; not for general code review; application code.

How do I install Kb in Claude Code?

Run `npx skills add compozy/kb --skill kb -a claude-code`. Or copy the skill folder (skills/kb in compozy/kb) into .claude/skills/kb in your project. Claude Code loads it when a task matches its description.

How do I install Kb in Codex?

Run `npx skills add compozy/kb --skill kb -a codex`. Or copy the skill folder (skills/kb in compozy/kb) into .agents/skills/kb in your project. Codex loads it when a task matches its description.

Can I use Kb in Cursor, Gemini CLI or GitHub Copilot?

Cursor, Gemini CLI, GitHub Copilot and OpenCode also load SKILL.md folders. With the skills CLI, run `npx skills add compozy/kb --skill kb -a cursor` (or -a gemini-cli, github-copilot or opencode for the others). To copy it by hand, put the folder in .cursor/skills/kb, .gemini/skills/kb, .github/skills/kb and .opencode/skills/kb in your project.

What does Kb need to run?

Going by SKILL.md and its folder, Kb needs the command-line tools its instructions call (npm) and credentials named OPENROUTER_API_KEY and FIRECRAWL_API_KEY. Our summary lists: Node.js; A credential in OPENROUTER_API_KEY; A credential in FIRECRAWL_API_KEY.

Does Kb access the network?

SKILL.md names 1 domain. As links in the text: github.com. This is read from the text; nothing was executed.

Is Kb safe to install?

Our automated static check of SKILL.md found no risky patterns, such as piping downloads into a shell, reading credential files or hidden Unicode. It is not a guarantee. Review the folder before installing.

What licence does Kb use?

Kb is published under the MIT licence (the repository's licence). It allows redistribution, so the full SKILL.md is shown on this page.

How many tokens does Kb use?

About 4.2k tokens (SKILL.md is roughly 17k characters). Agents keep only the skill's name and description in context until a task matches; then they load SKILL.md in full. Its references folder adds about 32k tokens, read only when the agent opens those files.

What are the alternatives to Kb?

Skills that share tags, products or a category with Kb: LLM Wiki (lewislulu/llm-wiki-skill, 655 stars), Knowledge Base Health Lint (jason-effi-lab/karpathy-llm-wiki-vault, 715 stars), LLM Wiki Knowledge Graph (Egonex-AI/Understand-Anything, 85k stars) and Wiki Builder (rohitg00/pro-workflow, 2.9k stars). The comparison table on this page puts their stars, adoption, token cost, safety result and licence side by side.

Who maintains Kb?

compozy (a GitHub organization) maintains it in compozy/kb, which has 110 GitHub stars. The repository was last updated on September 25, 2026.

Source: compozy/kb on GitHub. Facts on this page come from the repository at the commit we read; the author's words are quoted as theirs.