# Documentation


The platform runs one loop. You stand up the environment, run sessions, see
what happened, remember it, and measure the outcome. Each job below covers one
part of that loop. Start with the first job, then pick the job you have now.

## Stand Up the Platform (Teams Using Agents)

<div class="grid">

<a href="getting-started/">
<h3>Getting Started: Your First Trace</h3>
<p>Install the Gemba skill pack and the gemba-* command family, run one agent session, and read the NDJSON trace it captured.</p>
</a>

</div>

## Coordinate an Agent Team (Teams Using Agents)

<div class="grid">

<a href="coordinate-team/">
<h3>Coordinate an Agent Team</h3>
<p>Run a lead and N participant agents in one asynchronous session. Choose supervise, facilitate, or discuss. Pass messages with Ask, Answer, and Announce. One NDJSON trace records everything that happened.</p>
</a>

</div>

## Prove Agent Changes (Teams Using Agents)

<div class="grid">

<a href="prove-changes/">
<h3>Prove Agent Changes</h3>
<p>Reproducible evidence that agent changes improved outcomes, from the eval session through the trace analysis.</p>
</a>

<a href="prove-changes/run-eval/">
<h3>Run an Eval</h3>
<p>Run an agent-as-judge eval in CI and get a traceable verdict on whether an agent change improved outcomes.</p>
</a>

<a href="prove-changes/run-benchmark/">
<h3>Run a Benchmark</h3>
<p>Prove a skill-pack change improved coding outcomes. Run a task family across N runs, grade with hidden tests, and report pass@k.</p>
</a>

<a href="prove-changes/run-benchmark/ci-workflow/">
<h3>Automate with GitHub Actions</h3>
<p>Run gemba-benchmark in CI with the forwardimpact/gemba-benchmark composite action. You get step summaries, artifact upload, and PR-triggered benchmarks.</p>
</a>

<a href="prove-changes/trace-analysis/">
<h3>Analyze Traces</h3>
<p>See exactly what an agent did and why. Download traces, query turns, filter by tool or error, and measure token cost.</p>
</a>

</div>

## Operate a Predictable Agent Team (Teams Using Agents)

<div class="grid">

<a href="predictable-team/">
<h3>Set Up Persistent Memory and Metrics</h3>
<p>Give your agent team persistent memory and real signal detection with wiki-backed state and XmR control charts. Get evidence that agents act on changes. They do not act on noise.</p>
</a>

<a href="predictable-team/wiki-operations/">
<h3>Send a Memo or Update a Storyboard</h3>
<p>Communicate across your agent team and keep storyboards current. You do not manage the wiki infrastructure yourself.</p>
</a>

<a href="predictable-team/wiki-integrity/">
<h3>Audit and Auto-Fix the Wiki</h3>
<p>Keep the wiki valid against a declarative rule catalogue. Auto-fix what is safely fixable. Surface the rest for a human, so stale memory never poisons coordination.</p>
</a>

<a href="predictable-team/collision-ledger/">
<h3>Allocate Collision-Ledger Entries for Parallel Work</h3>
<p>Assign stable ids to parallel work without merge collisions. An append-only issue thread anchors every id. The ledger page gets a projection only when you rebuild.</p>
</a>

<a href="predictable-team/xmr-analysis/">
<h3>Chart a Metric and Check Variation</h3>
<p>Know whether a metric changed or only varied. Natural process limits and Wheeler's detection rules separate signal from noise.</p>
</a>

</div>

The platform ships commands and CI actions. It adds no importable API of its
own. When you want the components instead of the commands, import the runtime
libraries directly. See the
[library catalog](https://github.com/forwardimpact/monorepo/blob/main/libraries/README.md#catalog)
for that API surface.

[Kata](https://www.kata.team/) is the reference tenant. It runs this platform
every day. The Kata site documents the agent-team practice that sits on top of
the loop.