> ## Documentation Index
> Fetch the complete documentation index at: https://docs.woes.dev/llms.txt
> Use this file to discover all available pages before exploring further.

# Context Learning

> Use context learning to auto-tune an agent from its attached sources.

# Context Learning

Context learning analyzes an agent's attached sources and creates a context-aware operating brief. It helps the agent answer from the actual domains, endpoints, and constraints it is responsible for.

## What Learning Produces

Context learning can produce:

* Markdown operating brief.
* Summary of what the agent covers.
* Domains and capabilities.
* Per-source focus notes.
* Coverage notes.
* Cautions and guardrails.
* Tuned persona.
* Tuned runbook instructions.
* Greeting message.
* Confidence threshold suggestions.
* Custom guardrails.

## Statuses

| Status  | Meaning                                                                |
| ------- | ---------------------------------------------------------------------- |
| Idle    | No learning run is pending or ready.                                   |
| Pending | The agent has context and is waiting for a learning run.               |
| Ready   | A learned profile is available.                                        |
| Failed  | The learning run failed; review the operator-readable error and retry. |

## Run Learning

<Steps>
  <Step title="Attach sources">
    Add source context to the agent.
  </Step>

  <Step title="Open Context intelligence">
    Use the Context intelligence button in the agent overview.
  </Step>

  <Step title="Tune or re-tune">
    Run learning. Woes samples attached sources and generates a grounded operating brief.
  </Step>

  <Step title="Review before relying on it">
    Confirm the brief accurately describes the agent's domain and boundaries.
  </Step>
</Steps>

<Warning>
  Context learning should reinforce grounding. It should not authorize the agent to answer outside attached source evidence.
</Warning>