Guide · Agent skill
How to use the Knowledge Agent agent skill
Updated · written with AI from the project's own documentation and checked against our catalogue data
The knowledge-agent skill lets an agent turn your claude-mem observation history into named, filtered collections called corpora, then answer your questions from them in a dedicated AI session. It is for developers and teams who already use claude-mem and want a focused reference on one topic instead of searching everything. After this guide you will know what the skill does, how to install it, how to trigger it with plain requests, and what to check before sending history into an AI session.
Knowledge Agentthedotmack/claude-mem · Apache-2.0 Download the skill Quick start
- 1Install claude-mem history. Confirm claude-mem is installed and has an existing observation history, and that the claude-mem tool functions are available to the agent.
- 2Install the skill. Download the ZIP from this page and unzip it into ~/.claude/skills/ or a project's .claude/skills/ folder, or upload it under Settings → Capabilities → Skills in the Claude apps.
- 3Build a focused corpus. Ask the agent to build a corpus with a name, description and filters such as project, types, concepts, files, query or date range.
- 4Prime the corpus. Load the corpus into an AI session with prime_corpus. Large corpora take a moment.
- 5Ask questions. Use query_corpus with your question and follow up in the same session.
- 6Rebuild and reprime when needed. After new observations are added, run rebuild_corpus and then reprime_corpus to load fresh knowledge.
What does the knowledge-agent skill make the agent do?
The skill builds knowledge agents: filtered corpora of your claude-mem observations (notes the memory tool recorded about past work). You build a corpus, prime it so its knowledge is loaded into an AI session, then ask questions conversationally.
Examples of corpora from the source include 'everything about hooks', 'all decisions from the last month', and 'all bugfixes for the worker service'.
Install the skill from the ZIP or the Claude apps
Download the ZIP from the skill's page on this site. Then choose one of these routes:
Unzip it into ~/.claude/skills/ to make it available to your user account, or into a project's .claude/skills/ folder to limit it to that project.
Or upload it in the Claude apps under Settings → Capabilities → Skills.
- Requires claude-mem installed with an existing observation history.
- The claude-mem tool functions build_corpus, prime_corpus, query_corpus, list_corpora, rebuild_corpus and reprime_corpus must be available to the agent.
How do you trigger the skill with natural requests?
The skill's description says to use it when you want to create focused 'brains' from your observation history, ask about past work patterns, or compile expertise on a specific topic. You do not need to type the tool names yourself; phrase the request in plain language, such as asking the agent to build a knowledge base about hooks from your claude-mem history.
The workflow the agent follows has four steps: build a corpus, prime it, query it, and list corpora.
Step 1: Build a corpus with filters
A corpus is created with build_corpus, a name, a description and optional filters. The source lists these filters: project (project name), types (comma-separated from decision, bugfix, feature, refactor, discovery, change), concepts (comma-separated concept tags), files (comma-separated file paths, prefix match), query (semantic search query), dateStart and dateEnd (ISO date range), and limit (maximum observations, default 500).
- Focused corpora work best: 'hooks architecture' beats 'everything ever'.
build_corpus name="hooks-expertise" description="Everything about the hooks lifecycle" project="claude-mem" concepts="hooks" limit=500Step 2: Prime and query the corpus
Priming with prime_corpus creates an AI session loaded with the corpus knowledge. The source notes this takes a moment for large corpora. Once primed, ask questions with query_corpus. Follow-up questions keep their context within the session, and the session persists across queries.
prime_corpus name="hooks-expertise"
query_corpus name="hooks-expertise" question="What are the 5 lifecycle hooks and when does each fire?"Keep corpora current: list, rebuild and reprime
Use list_corpora to see all corpora with stats and priming status. When new observations are added, run rebuild_corpus to refresh the corpus, then reprime_corpus to load the updated knowledge. Reprime clears prior Q&A context and reloads the corpus into a new session, which is also useful when the conversation drifts.
list_corpora
rebuild_corpus name="hooks-expertise"
reprime_corpus name="hooks-expertise"Safety notes: scripts and network use
The skill runs no scripts. It does use the network: priming a corpus sends that observation history into an AI session, which may be an external service. Review what a corpus includes before priming it, especially if your observations are confidential.
Answers come only from recorded observations, which may be stale or incomplete, so do not treat them as authoritative facts about the outside world.
- Skip priming corpora that contain confidential observations until you have reviewed what is included.
What people use it for
- Reviewing what was decided about hooks in a long-running project.
- Finding which bugfixes touched the worker service.
- Building a focused reference on one area of a project from its recorded history.
- Summarising all decisions from a recent date range using dateStart and dateEnd filters.
Frequently asked questions
Is the knowledge-agent skill free?
The source material for this page does not state pricing. The skill is published under the Apache-2.0 licence. Check the licence file in the repository for the terms that apply to you.
What do I need before installing knowledge-agent?
You need claude-mem installed with an existing observation history, and the claude-mem tool functions build_corpus, prime_corpus, query_corpus, list_corpora, rebuild_corpus and reprime_corpus available to the agent. Without observations there is nothing to build a corpus from.
Where do I install the knowledge-agent skill?
Unzip the downloaded ZIP into ~/.claude/skills/ for your user account, or into a project's .claude/skills/ folder for one project. You can also upload it in the Claude apps under Settings → Capabilities → Skills.
Does knowledge-agent run scripts or use the internet?
The skill runs no scripts. It does use the network, because priming a corpus sends its observation history into an AI session, which may be an external service.
Can I use knowledge-agent without claude-mem?
No. The skill works only on claude-mem observation history and has no data of its own, so it needs claude-mem and recorded observations.
How is this different from searching my history directly?
Instead of searching everything, you build a focused corpus by project, observation type, concept tags, file paths, search query or date range, then ask follow-up questions that build on earlier answers in the same session.
New agent skills, repos and AI tools — every week
The best new skills and open-source projects, the fastest-growing repos, and one free prompt. Five minutes to read. Unsubscribe anytime.
Our verdict
Use knowledge-agent if you already use claude-mem and want to review your own past decisions and bugfixes on a specific topic, or to build a focused reference from a long-running project's history. Pick something else if you do not use claude-mem, need authoritative outside facts, or have confidential observations you are not willing to send to an AI session.
Download the skill