diff --git a/README.md b/README.md
index 80e869f..98979ae 100644
--- a/README.md
+++ b/README.md
@@ -83,6 +83,31 @@ All routes require `manage_options` (filter: `asa_capability`):
 - `GET /wp-json/asa/v1/servers` — MCP server inventory + adapter state
 - `GET /wp-json/asa/v1/export` — report as a download attachment
 
+### Fleet summary
+
+The report carries a compact `fleet_summary` — a one-glance digest of this
+site's agent-surface posture, meant for fleet tooling that needs "how exposed is
+this one site?" without the full per-ability report:
+
+```json
+"fleet_summary": {
+  "sev": "crit",
+  "highest_severity": "critical",
+  "agent_reachable": 19,
+  "agent_reachable_write": 3,
+  "findings_by_severity": { "critical": 3, "high": 7, "medium": 9, "low": 6, "info": 1 },
+  "analysis_errors": 0
+}
+```
+
+`sev` (`ok`/`warn`/`crit`) is derived conservatively from the highest finding
+severity present (critical/high → `crit`, medium/low → `warn`, info or none →
+`ok`) and, like the rest of the report, means "no issues detected", never
+"safe". It is a pure projection of the audit — no extra analysis — and is
+consumed as a fleet-health signal (e.g. by a MainWP Fleet Health Monitor via its
+`fhm_extra_signals` filter). Treat the field as optional: a report cached before
+an upgrade may briefly omit it.
+
 ## WP-CLI & CI
 
 The plugin registers `wp asa audit` (read-only, like everything else):
diff --git a/src/Report/class-audit-runner.php b/src/Report/class-audit-runner.php
index 1505821..aab8d0d 100644
--- a/src/Report/class-audit-runner.php
+++ b/src/Report/class-audit-runner.php
@@ -140,21 +140,108 @@ private function compute() {
 		}
 
 		return [
-			'generated_at' => gmdate( 'c' ),
-			'from_cache'   => false,
-			'environment'  => [
+			'generated_at'  => gmdate( 'c' ),
+			'from_cache'    => false,
+			'environment'   => [
 				'plugin_version' => ASA_VERSION,
 				'wp_version'     => is_string( $wp_version ) ? $wp_version : '',
 				'php_version'    => PHP_VERSION,
 				'mcp_adapter'    => $adapter_state,
 			],
-			'summary'      => $summary,
-			'coverage'     => $this->coverage_notes(),
-			'servers'      => $public_servers,
-			'abilities'    => $abilities,
+			'summary'       => $summary,
+			'fleet_summary' => $this->fleet_summary( $summary ),
+			'coverage'      => $this->coverage_notes(),
+			'servers'       => $public_servers,
+			'abilities'     => $abilities,
 		];
 	}
 
+	/**
+	 * A compact, transport-agnostic digest of this site's agent-surface posture.
+	 *
+	 * This is the stable contract that fleet consumers read — a MainWP child
+	 * reporter, a CI gate, or anything that needs "how bad is this one site?"
+	 * without the full per-ability report. It is a pure projection of the
+	 * already-computed $summary: no new analysis, no I/O, no state. Keeping it
+	 * small is deliberate — a fleet payload should carry counts and a grade, not
+	 * ability descriptors (which would enlarge what leaves the site and duplicate
+	 * the full report).
+	 *
+	 * `sev` speaks the Fleet Health Monitor vocabulary (`ok|warn|crit`) so an
+	 * `fhm_extra_signals` consumer can hand it straight to the grade. Note this
+	 * grade reflects the *highest finding severity present*, which is a smell
+	 * signal, not a safety proof — consistent with §5 (a green tile is "no issues
+	 * detected", never "safe").
+	 *
+	 * @param array<string, mixed> $summary The computed report summary.
+	 * @return array<string, mixed> The fleet digest.
+	 */
+	private function fleet_summary( array $summary ) {
+		$by_severity = isset( $summary['findings_by_severity'] ) && is_array( $summary['findings_by_severity'] )
+			? $summary['findings_by_severity']
+			: [];
+
+		// Highest severity that actually has ≥1 finding (most severe first).
+		$highest = null;
+		foreach ( Finding::SEVERITY_ORDER as $severity ) {
+			if ( ! empty( $by_severity[ $severity ] ) ) {
+				$highest = $severity;
+				break;
+			}
+		}
+
+		$by_severity_ints = [];
+		foreach ( $by_severity as $severity => $count ) {
+			$by_severity_ints[ (string) $severity ] = $this->int_of( $count );
+		}
+
+		return [
+			'sev'                   => $this->derive_sev( $highest ),
+			'highest_severity'      => $highest, // null when no findings at all.
+			'agent_reachable'       => $this->int_of( isset( $summary['agent_reachable'] ) ? $summary['agent_reachable'] : 0 ),
+			'agent_reachable_write' => $this->int_of( isset( $summary['agent_reachable_write'] ) ? $summary['agent_reachable_write'] : 0 ),
+			'findings_by_severity'  => $by_severity_ints,
+			'analysis_errors'       => $this->int_of( isset( $summary['analysis_errors'] ) ? $summary['analysis_errors'] : 0 ),
+		];
+	}
+
+	/**
+	 * Narrow a mixed value to a non-negative int for the digest.
+	 *
+	 * The summary is typed `array<string,mixed>`, so its counts are `mixed` at
+	 * the type level; this proves the int for both the reader and the analyzer.
+	 *
+	 * @param mixed $value Any value; only numerics contribute a count.
+	 * @return int The value as an int, or 0 when not numeric.
+	 */
+	private function int_of( $value ) {
+		return is_numeric( $value ) ? (int) $value : 0;
+	}
+
+	/**
+	 * Map the highest finding severity to the Fleet Health `ok|warn|crit` grade.
+	 *
+	 * Critical/high map to crit (an agent-reachable write or missing gate is a
+	 * fleet alarm); medium/low map to warn (a smell worth surfacing, not an
+	 * alarm); info or no findings map to ok. Deliberately conservative: info-only
+	 * sites (e.g. a bare ASA011 meta-tool awareness note) do not raise a grade.
+	 *
+	 * @param string|null $highest Highest present severity, or null.
+	 * @return string One of 'ok'|'warn'|'crit'.
+	 */
+	private function derive_sev( $highest ) {
+		switch ( $highest ) {
+			case Finding::SEVERITY_CRITICAL:
+			case Finding::SEVERITY_HIGH:
+				return 'crit';
+			case Finding::SEVERITY_MEDIUM:
+			case Finding::SEVERITY_LOW:
+				return 'warn';
+			default:
+				return 'ok';
+		}
+	}
+
 	/**
 	 * Standing statements about what this audit does NOT cover.
 	 *
diff --git a/tests/Integration/report-integration-test.php b/tests/Integration/report-integration-test.php
index 40981a6..a296741 100644
--- a/tests/Integration/report-integration-test.php
+++ b/tests/Integration/report-integration-test.php
@@ -139,4 +139,25 @@ public function test_summary_counts_are_consistent() {
 		$this->assertGreaterThanOrEqual( 1, $summary['findings_by_severity']['critical'] );
 		$this->assertGreaterThanOrEqual( 1, $summary['agent_reachable_write'] );
 	}
+
+	public function test_fleet_summary_is_a_faithful_projection_of_the_summary() {
+		$this->assertArrayHasKey( 'fleet_summary', self::$report );
+		$fleet   = self::$report['fleet_summary'];
+		$summary = self::$report['summary'];
+
+		// Shape: the compact digest a fleet consumer reads.
+		foreach ( [ 'sev', 'highest_severity', 'agent_reachable', 'agent_reachable_write', 'findings_by_severity', 'analysis_errors' ] as $key ) {
+			$this->assertArrayHasKey( $key, $fleet, "fleet_summary missing $key" );
+		}
+
+		// It is a pure projection — counts equal the source summary's.
+		$this->assertSame( (int) $summary['agent_reachable'], $fleet['agent_reachable'] );
+		$this->assertSame( (int) $summary['agent_reachable_write'], $fleet['agent_reachable_write'] );
+		$this->assertSame( array_map( 'intval', $summary['findings_by_severity'] ), $fleet['findings_by_severity'] );
+
+		// The fixtures produce ≥1 critical finding, so the grade must be 'crit'
+		// and the highest severity 'critical'.
+		$this->assertSame( 'critical', $fleet['highest_severity'] );
+		$this->assertSame( 'crit', $fleet['sev'] );
+	}
 }
diff --git a/tests/Unit/fleet-summary-test.php b/tests/Unit/fleet-summary-test.php
new file mode 100644
index 0000000..d8a5966
--- /dev/null
+++ b/tests/Unit/fleet-summary-test.php
@@ -0,0 +1,112 @@
+<?php
+/**
+ * Fleet_Summary tests: the compact fleet digest and its severity→grade mapping.
+ *
+ * These are pure-PHP unit tests. Audit_Runner's fleet_summary()/derive_sev() do
+ * not touch WordPress (no I/O, no globals), so we exercise them via reflection
+ * without a WP install — only run()/compute() need WordPress.
+ *
+ * @package ASA
+ */
+
+use ASA\Report\Audit_Runner;
+use ASA\Model\Finding;
+
+/**
+ * @covers \ASA\Report\Audit_Runner
+ */
+class Fleet_Summary_Test extends ASA_TestCase {
+
+	/**
+	 * Invoke a private method on a fresh Audit_Runner.
+	 *
+	 * @param string  $method Method name.
+	 * @param mixed[] $args   Arguments.
+	 * @return mixed The return value.
+	 */
+	private function invoke( $method, array $args ) {
+		$runner = new Audit_Runner();
+		$ref    = new ReflectionMethod( Audit_Runner::class, $method );
+		$ref->setAccessible( true );
+		return $ref->invokeArgs( $runner, $args );
+	}
+
+	/**
+	 * A summary array with the given findings-by-severity counts.
+	 *
+	 * @param array<string,int> $by_severity Severity => count.
+	 * @param int               $reachable_write Agent-reachable writes.
+	 * @return array<string,mixed>
+	 */
+	private function summary( array $by_severity, $reachable_write = 0 ) {
+		return [
+			'agent_reachable'       => 3,
+			'agent_reachable_write' => $reachable_write,
+			'analysis_errors'       => 0,
+			'findings_by_severity'  => array_merge(
+				array_fill_keys( Finding::SEVERITY_ORDER, 0 ),
+				$by_severity
+			),
+		];
+	}
+
+	public function test_critical_finding_grades_crit() {
+		$fleet = $this->invoke( 'fleet_summary', [ $this->summary( [ 'critical' => 1 ] ) ] );
+
+		$this->assertSame( 'crit', $fleet['sev'] );
+		$this->assertSame( 'critical', $fleet['highest_severity'] );
+	}
+
+	public function test_high_finding_grades_crit() {
+		$fleet = $this->invoke( 'fleet_summary', [ $this->summary( [ 'high' => 2 ] ) ] );
+
+		$this->assertSame( 'crit', $fleet['sev'] );
+		$this->assertSame( 'high', $fleet['highest_severity'] );
+	}
+
+	public function test_medium_and_low_grade_warn() {
+		$this->assertSame( 'warn', $this->invoke( 'fleet_summary', [ $this->summary( [ 'medium' => 1 ] ) ] )['sev'] );
+		$this->assertSame( 'warn', $this->invoke( 'fleet_summary', [ $this->summary( [ 'low' => 1 ] ) ] )['sev'] );
+	}
+
+	public function test_info_only_grades_ok() {
+		// An info-only site (e.g. a bare ASA011 awareness note) is NOT a fleet alarm.
+		$fleet = $this->invoke( 'fleet_summary', [ $this->summary( [ 'info' => 3 ] ) ] );
+
+		$this->assertSame( 'ok', $fleet['sev'] );
+		$this->assertSame( 'info', $fleet['highest_severity'] );
+	}
+
+	public function test_no_findings_grades_ok_with_null_highest() {
+		$fleet = $this->invoke( 'fleet_summary', [ $this->summary( [] ) ] );
+
+		$this->assertSame( 'ok', $fleet['sev'] );
+		$this->assertNull( $fleet['highest_severity'] );
+	}
+
+	public function test_highest_severity_wins_when_multiple_present() {
+		// low + critical together → the grade reflects the worst present.
+		$fleet = $this->invoke( 'fleet_summary', [ $this->summary( [ 'low' => 5, 'critical' => 1 ] ) ] );
+
+		$this->assertSame( 'critical', $fleet['highest_severity'] );
+		$this->assertSame( 'crit', $fleet['sev'] );
+	}
+
+	public function test_projection_carries_counts_verbatim() {
+		$fleet = $this->invoke( 'fleet_summary', [ $this->summary( [ 'high' => 1 ], 4 ) ] );
+
+		$this->assertSame( 3, $fleet['agent_reachable'] );
+		$this->assertSame( 4, $fleet['agent_reachable_write'] );
+		$this->assertSame( 0, $fleet['analysis_errors'] );
+		$this->assertSame( 1, $fleet['findings_by_severity']['high'] );
+	}
+
+	public function test_tolerates_a_summary_missing_findings_by_severity() {
+		// Defensive: a malformed summary must not fatal — grade defaults to ok.
+		$fleet = $this->invoke( 'fleet_summary', [ [ 'agent_reachable' => 0 ] ] );
+
+		$this->assertSame( 'ok', $fleet['sev'] );
+		$this->assertNull( $fleet['highest_severity'] );
+		$this->assertSame( [], $fleet['findings_by_severity'] );
+	}
+}