<divclass="wrap"><header><divclass="eyebrow"><span>canon-autonomy-lanes</span><spanclass="stat">accepted · accepted-1</span><span>the-custodian</span><span>reviewed 2026-08-31</span><span>generated from canonical source — do not edit</span></div><h1>Autonomy Lanes (Fleet) v0.1</h1><pclass="sub">Source: <code>the-custodian · canon/standards/autonomy-lanes_v0.1.md · f9435cd605cc5b3cb0f2e957ce6287d9f3129aac</code></p><pclass="sub">Review due: 2027-02-28</p></header><divclass="layout"><navclass="rail"aria-label="Sections"><ol><li><ahref="#the-rule"><spanclass="n">·</span>The rule</a></li><li><ahref="#lanes"><spanclass="n">·</span>Lanes</a></li><li><ahref="#lane-as-a-spine-field"><spanclass="n">·</span>Lane as a spine field</a></li><li><ahref="#approval-packages-yellow-and-above"><spanclass="n">·</span>Approval packages (yellow and above)</a></li><li><ahref="#human-attention-is-a-wip-limited-workstation"><spanclass="n">·</span>Human attention is a WIP-limited workstation</a></li><li><ahref="#raising-autonomy"><spanclass="n">·</span>Raising autonomy</a></li><li><ahref="#references"><spanclass="n">·</span>References</a></li></ol></nav><main><divclass="rule-quote"><p>Ratified by founder 2026-07-20 (CUST-WP-0060-T01, hub decision f4640f9e). Promotes the binky-control AutonomyPolicy lane model to fleet canon, unchanged in substance. binky-control's <code>AutonomyPolicy.md</code> remains the company-level policy instance; this standard makes the lane vocabulary and enforcement rules fleet-wide.</p></div>
<divclass="rule-quote"><p>If the responsible human is unavailable, the system must either continue safely, prepare the next decision, or explicitly defer with evidence. It must not silently idle.</p></div>
<p>Valid states: <code>proceeding</code> | <code>prepared for review</code> | <code>deferred by policy</code>. Invalid state: <code>waiting because unsure</code>. Uncertainty triggers research, comparison, preparation, or risk classification — not paralysis. "Ask the human" is never the default.</p>
</section>
<sectionid="lanes"><h2>Lanes</h2>
<divclass="scroll"><table><thead><tr><th>Lane</th><th>Name</th><th>Agent may</th><th>Human involvement</th></tr></thead><tbody><tr><td><strong>green</strong></td><td>Safe autonomous</td><td>research, summarize, classify, draft, organize, test, prepare</td><td>none (human-on-the-loop via logs)</td></tr><tr><td><strong>blue</strong></td><td>Reversible execution</td><td>internal docs, backlog updates, local scripts, reports, local commits</td><td>none, but logged</td></tr><tr><td><strong>yellow</strong></td><td>Deferred approval</td><td>prepare external emails, invoices, offers, releases, payments</td><td>async approval via decision records</td></tr><tr><td><strong>orange</strong></td><td>Time-window approval</td><td>customer/legal/accounting/banking actions needing business hours</td><td>engagement records, batched into command days</td></tr><tr><td><strong>red</strong></td><td>Human-only</td><td>signing, strategic commitments, bank/credential changes, sensitive legal/HR/security</td><td>always; agents prepare packages only</td></tr></tbody></table></div>
<p>Optimization target: <strong>maximize green+blue throughput, queue yellow, batch orange, protect red from automation.</strong> Risk-tiered gating (in-the-loop for irreversible actions, on-the-loop for recoverable ones) is the only oversight pattern that scales.</p>
</section>
<sectionid="lane-as-a-spine-field"><h2>Lane as a spine field</h2>
<p><code>lane</code> is a <strong>mandatory field on every work record</strong> (see <code>work-record-types_v0.1.md</code>) and on every agent instance manifest. The chain is closed end-to-end:</p>
→ ops-warden credential lane (catalog access)</pre>
<p>Declaring a lane an item does not hold is a policy violation the harness must catch (tool profile refuses the run).</p>
</section>
<sectionid="approval-packages-yellow-and-above"><h2>Approval packages (yellow and above)</h2>
<p>An agent hitting a lane boundary prepares a complete decision package, never a bare question: recommended action, reason, evidence links, risk level, reversibility, deadline (<code>needed_by</code>), attention cost, fallback if no response, options.</p>
</section>
<sectionid="human-attention-is-a-wip-limited-workstation"><h2>Human attention is a WIP-limited workstation</h2>
<p>The responsible human's capacity is modeled like any lane capacity: attention budget (minutes/day, max decisions/day, max deep topics/day) and availability as first-class state (<code>available | evening_window | unavailable | office_hour_day</code>), with an explicit escalation allow-list when unavailable (critical security incident, payment failure, legal deadline). Yellow routes to the decision view, orange to the engagement view, both subject to that capacity.</p>
<p><strong>Time-boxing is hub-owned:</strong><code>needed_by</code> / <code>fallback_if_no_response</code> windows are watched by the hub → NATS → activity-core; a lapsed window raises the escalation or executes the declared fallback — never a human re-reading a file.</p>
<p>Autonomy per area is raised only after the system proves reliable there, by explicit human decision recorded as a decision work record. Lane assignments are auditable history; lowering a lane needs no approval.</p>
