Openclaw Skills for Agent KB helps you create, maintain, validate, and optimize a repository-local knowledge base for coding agents.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install agent-context-kb
2. Manual Installation
Copy the skill folder to one of these locations
~/.openclaw/skills/ <project>/skills/ Priority: Workspace > Local > Bundled
3. Prompt Installation
Copy this prompt to OpenClaw to install it automatically.
Help me install agent-context-kb using Clawhub. If Clawhub is not installed, install it first (npm i -g clawhub).
Prefer to download?
Get the raw skill files in a ZIP archive.
What is Agent KB?
Openclaw Skills for Agent KB is a repository-scoped maintenance skill for managing a .agent-kb/ knowledge base designed specifically for coding agents. It focuses on durable, agent-facing project memory: scaffolding the KB, refreshing runtime protocol files, validating structure and routes, compiling inbox notes into stable topics, and trimming stale or redundant content.
Unlike general documentation tooling, this skill is built around a strict separation of concerns between protocol, canon, and maintenance. Openclaw Skills ensures the KB stays compact, reachable, and operational so agents can reliably find the right project knowledge without duplicating human documentation.
Agent KB Use Cases
- Setting up a new
.agent-kb/repository for agent memory and project continuity - Refreshing the KB runtime protocol or scaffold after skill updates
- Validating KB edits, route definitions, and link reachability before finishing work
- Capturing durable facts when the best permanent file target is unclear
- Compiling inbox notes into valid topic files once the destination becomes clear
- Trimming stale, redundant, or overly large KB content to keep the knowledge base lean
- Recording current focus, milestones, and open questions in
.agent-kb/plans/current.md - Managing different versioning modes: nested, shared, or local
- Backfilling usage stats and transcript reads when available
How Agent KB Works
- Initialize or upgrade the repository KB using the bundled Python script from the skill directory.
- Choose the correct versioning mode for the repository: nested by default, or shared/local when requested.
- Add durable knowledge through notes when the stable destination is unclear, or write directly to the appropriate topic file when it is known.
- Compile inbox notes into topic files to merge temporary captures into lasting knowledge.
- Validate the KB after every meaningful edit to catch broken routes, links, or protocol issues.
- Trim stale or repetitive information when the KB grows too large, using dry-run output first and
--writeonly for deterministic cleanup. - Use stats to inspect KB health and transcript-backed usage patterns, then paste the full ASCII output verbatim when required.
Agent KB Setup
- Open the skill directory that contains the bundled KB management script.
- Run initialization against the target repository root.
- Validate the scaffold and any generated runtime protocol files.
- If the scaffold is empty after init, present the one-time distillation pass to the user and proceed only if accepted.
- Configure versioning mode as needed: nested default, shared, or local.
- Keep
.agent-kb/.gitignoreand.agent-kb/.kb-meta.yamlaligned with the chosen mode. - Regenerate the route map after route changes.
python3 scripts/agent_kb.py init --root /path/to/repo
python3 scripts/agent_kb.py upgrade --root /path/to/repo
python3 scripts/agent_kb.py validate --root /path/to/repo
python3 scripts/agent_kb.py note --root /path/to/repo --title "Auth session note" --target decisions/active/auth-storage.md --body "Durable fact."
python3 scripts/agent_kb.py compile --root /path/to/repo
python3 scripts/agent_kb.py trim --root /path/to/repo
python3 scripts/agent_kb.py trim --root /path/to/repo --write
python3 scripts/agent_kb.py stats --root /path/to/repo
Agent KB Data Schema & Taxonomy
Openclaw Skills for Agent KB organizes repository knowledge under a .agent-kb/ directory with a strict topical structure and metadata-driven routing.
| Path / File | Purpose | Notes |
|---|---|---|
.agent-kb/start.md |
Canonical durable-knowledge rules | Defines what counts as durable knowledge and the do/don't canon |
.agent-kb/plans/current.md |
Lightweight continuity plan | Stores current focus, milestones, next moves, and open questions |
.agent-kb/routes.yaml |
Canonical route source | Maps tasks to the files agents should read first |
.agent-kb/map.md |
Generated route map | Regenerated when routes change |
.agent-kb/.kb-meta.yaml |
Versioning metadata | Records nested/shared/local mode |
.agent-kb/.gitignore |
Privacy and VCS control | Always excludes .log/ from version control |
Topic files under decisions/, workflows/, etc. |
Durable project knowledge | Must remain reachable from routes or Markdown links |
Route taxonomy
id: stable route identifiertask: human-readable purpose for the routeread_first: exactly one primary filealso_consider: up to two additional files
Maintenance taxonomy
init: scaffold and protocol generationupgrade: conservative refresh of generated filesvalidate: integrity and link checksnote: durable fact capture when destination is uncertaincompile: merge inbox notes into stable topicstrim: reduce redundancy and stalenessstats: report KB health and transcript-derived metrics
Agent KB Advanced Features
- Support for nested, shared, and local versioning modes to fit personal, team, or repo-local workflows
- Conservative
upgradebehavior to refresh generated scaffold files without unnecessary churn - Route validation that enforces reachability and link integrity across the KB
- Inbox-to-topic compilation for turning temporary notes into durable project memory
- Controlled trimming workflow that favors non-redundant, genuinely compact knowledge over aggressive deletion
statsoutput designed for verbatim reporting, including ASCII charts as the deliverable- Explicit separation of trigger, canon, and skill layers to prevent protocol duplication
- Route map regeneration to keep navigation aligned with evolving project knowledge
SKILL.md
Loading
Related Openclaw Skills
Karakeep Note Capture
Karakeep Note Capture lets Openclaw save durable, human-readable notes into Karakeep for later discovery outside chat and local memory.
PriceWin Hotel Price Comparison
PriceWin Hotel Price Comparison helps Openclaw Skills agents compare live hotel rates across major OTAs and OpenTravel to find the cheapest bookable option.
PriceWin Flight Search
Openclaw Skills for PriceWin Flight Search lets agents compare live flight fares across major OTAs with a strict polling workflow and booking links.
PriceWin Hotel Deal Finder
PriceWin Hotel Deal Finder uses Openclaw Skills to compare live hotel prices across major OTAs and return the cheapest, best-value, and quality picks in one command.
Xueshu Search
Xueshu Search is an Openclaw Skills academic search workflow for multilingual paper discovery, citation tracing, DOI lookup, BibTeX export, and AI summaries.
Agent Design Advisor
Openclaw Skills helps you decide whether a workflow is suitable for an AI agent and returns practical, best-practice design recommendations.
Featured*
