diff --git a/crates/wright-cli/src/cli.rs b/crates/wright-cli/src/cli.rs
index 63a1604..a9f0b7e 100644
--- a/crates/wright-cli/src/cli.rs
+++ b/crates/wright-cli/src/cli.rs
@@ -24,8 +24,8 @@ pub(crate) struct Cli {
 
 pub(crate) const LONG_ABOUT: &str = "Wright compiler and Workshop tooling CLI.
 
-Commands check correctness, report semantic facts, lint, inspect, compile, or reconstruct
-source through the typed wright-driver result envelope. `compile` and `convert`
+Commands check correctness, summarize semantic hotspots, lint, inspect exhaustive
+facts, compile, or reconstruct source through the typed wright-driver result envelope. `compile` and `convert`
 keep their source artifact stdout contracts; JSON mode prints only one
 wright-result/v1 envelope to stdout.
 
@@ -63,11 +63,11 @@ pub(crate) enum Command {
     Convert(ConvertArgs),
     /// Check frontend, project, semantic, and validation correctness.
     Check(CommonArgs),
-    /// Report semantic structure, symbol usage, and CFG measurements.
+    /// Summarize semantic structure, CFG hotspots, and cross-cutting state.
     Analyze(CommonArgs),
     /// Parse, lower, and report lint findings.
     Lint(LintArgs),
-    /// Parse, lower, and show the structural/semantic program model.
+    /// Parse, lower, and show exhaustive structural/semantic facts.
     Inspect(CommonArgs),
     /// Generate static shell completion from the command model.
     Completion(CompletionArgs),
diff --git a/crates/wright-cli/src/present.rs b/crates/wright-cli/src/present.rs
index 17da6e7..d56cdf3 100644
--- a/crates/wright-cli/src/present.rs
+++ b/crates/wright-cli/src/present.rs
@@ -259,9 +259,11 @@ fn render_verdict<T: serde::Serialize>(
             array_len(value, "/result/rules"),
         ),
         "analyze" => format!(
-            "{} symbol(s), {} rule measurement(s)",
+            "{} rule(s), {} symbol(s); ranked semantic report",
+            value
+                .pointer("/result/program")
+                .map_or(0, |program| count(program, "rules")),
             array_len(value, "/result/facts/symbols"),
-            array_len(value, "/result/facts/rules"),
         ),
         "inspect" => format!(
             "{} rule(s), {} symbol(s)",
@@ -527,6 +529,10 @@ fn render_convert<T: serde::Serialize>(envelope: &Envelope<T>) {
 
 fn render_analyze<T: serde::Serialize>(envelope: &Envelope<T>) {
     let value = serde_json::to_value(envelope).expect("envelope serializes");
+    let program = value
+        .pointer("/result/program")
+        .cloned()
+        .unwrap_or_default();
     let facts = value.pointer("/result/facts").cloned().unwrap_or_default();
     let symbols = facts
         .get("symbols")
@@ -538,61 +544,115 @@ fn render_analyze<T: serde::Serialize>(envelope: &Envelope<T>) {
         .and_then(serde_json::Value::as_array)
         .cloned()
         .unwrap_or_default();
-    println!("\nAnalysis details");
-    for symbol in &symbols {
-        let kind = symbol
-            .get("kind")
-            .and_then(serde_json::Value::as_str)
-            .unwrap_or("symbol");
-        let name = symbol
-            .get("name")
-            .and_then(serde_json::Value::as_str)
-            .unwrap_or("<unnamed>");
-        let usage = symbol.get("usage").cloned().unwrap_or_default();
-        println!(
-            "  {kind} {name}: reads {}, writes {}, calls {}, rules {}",
-            usage
-                .get("reads")
-                .and_then(serde_json::Value::as_u64)
-                .unwrap_or(0),
-            usage
-                .get("writes")
-                .and_then(serde_json::Value::as_u64)
-                .unwrap_or(0),
-            usage
-                .get("calls")
-                .and_then(serde_json::Value::as_u64)
-                .unwrap_or(0),
-            usage
-                .get("rules")
-                .and_then(serde_json::Value::as_u64)
-                .unwrap_or(0),
-        );
-    }
+    println!("\nProgram overview");
+    println!(
+        "  {} file(s), {} rule(s), {} global variable(s), {} player variable(s), {} subroutine(s)",
+        count(&program, "files"),
+        count(&program, "rules"),
+        count(&program, "globalVariables"),
+        count(&program, "playerVariables"),
+        count(&program, "subroutines"),
+    );
+    println!("  evidence: [static] parsed program inventory");
+
+    let mut total_blocks = 0;
+    let mut total_edges = 0;
+    let mut total_loops = 0;
+    let mut total_waits = 0;
+    let mut rule_hotspots = Vec::new();
     for rule in &rules {
-        let name = rule
-            .get("name")
-            .and_then(serde_json::Value::as_str)
-            .unwrap_or("<unnamed>");
         let flow = rule.get("controlFlow").cloned().unwrap_or_default();
-        println!(
-            "  rule {name}: {} blocks, {} edges, {} loop block(s), {} wait block(s)",
-            flow.get("blocks")
-                .and_then(serde_json::Value::as_u64)
-                .unwrap_or(0),
-            flow.get("edges")
-                .and_then(serde_json::Value::as_u64)
-                .unwrap_or(0),
-            flow.get("loopBlocks")
-                .and_then(serde_json::Value::as_u64)
-                .unwrap_or(0),
-            flow.get("waitBlocks")
-                .and_then(serde_json::Value::as_u64)
-                .unwrap_or(0),
-        );
+        let blocks = count(&flow, "blocks");
+        let edges = count(&flow, "edges");
+        let loops = count(&flow, "loopBlocks");
+        let waits = count(&flow, "waitBlocks");
+        total_blocks += blocks;
+        total_edges += edges;
+        total_loops += loops;
+        total_waits += waits;
+        rule_hotspots.push((
+            blocks + edges,
+            rule.get("name")
+                .and_then(serde_json::Value::as_str)
+                .unwrap_or("<unnamed>"),
+            blocks,
+            edges,
+            loops,
+            waits,
+        ));
+    }
+    rule_hotspots.sort_by(|left, right| right.0.cmp(&left.0).then_with(|| left.1.cmp(right.1)));
+    println!("\nControl-flow summary");
+    println!(
+        "  {total_blocks} blocks, {total_edges} edges, {total_loops} loop block(s), {total_waits} wait block(s)"
+    );
+    println!("  Top rules (heuristic ranking: blocks + edges; facts are [static])");
+    if rule_hotspots.is_empty() {
+        println!("    none");
+    } else {
+        for (_, name, blocks, edges, loops, waits) in rule_hotspots.iter().take(5) {
+            println!(
+                "    {name}: {blocks} blocks, {edges} edges, {loops} loop block(s), {waits} wait block(s)"
+            );
+        }
+    }
+
+    let mut coupled_symbols = symbols
+        .iter()
+        .filter(|symbol| {
+            matches!(
+                symbol.get("kind").and_then(serde_json::Value::as_str),
+                Some("globalVariable" | "playerVariable")
+            )
+        })
+        .map(|symbol| {
+            let usage = symbol.get("usage").cloned().unwrap_or_default();
+            let rules = count(&usage, "rules");
+            let reads = count(&usage, "reads");
+            let writes = count(&usage, "writes");
+            (
+                rules,
+                reads + writes,
+                symbol
+                    .get("kind")
+                    .and_then(serde_json::Value::as_str)
+                    .unwrap_or("variable"),
+                symbol
+                    .get("name")
+                    .and_then(serde_json::Value::as_str)
+                    .unwrap_or("<unnamed>"),
+                reads,
+                writes,
+            )
+        })
+        .collect::<Vec<_>>();
+    coupled_symbols.sort_by(|left, right| {
+        right
+            .0
+            .cmp(&left.0)
+            .then_with(|| right.1.cmp(&left.1))
+            .then_with(|| left.3.cmp(right.3))
+    });
+    println!("\nState and coupling");
+    println!("  Top variables (heuristic ranking: rules touched, then reads + writes)");
+    if coupled_symbols.is_empty() {
+        println!("    none");
+    } else {
+        for (rules, _, kind, name, reads, writes) in coupled_symbols.iter().take(5) {
+            println!(
+                "    {kind} {name}: {rules} rule(s), {reads} read(s), {writes} write(s) [static]"
+            );
+        }
     }
 }
 
+fn count(value: &serde_json::Value, key: &str) -> usize {
+    value
+        .get(key)
+        .and_then(serde_json::Value::as_u64)
+        .unwrap_or(0) as usize
+}
+
 fn render_lint<T: serde::Serialize>(envelope: &Envelope<T>) {
     let value = serde_json::to_value(envelope).expect("envelope serializes");
     let findings = value
diff --git a/crates/wright-cli/tests/cli.rs b/crates/wright-cli/tests/cli.rs
index 317a2c7..c234253 100644
--- a/crates/wright-cli/tests/cli.rs
+++ b/crates/wright-cli/tests/cli.rs
@@ -168,7 +168,7 @@ fn terminal_renderer_uses_command_specific_hierarchy() {
     for (command, heading, detail) in [
         ("check", "PASS check", "diagnostic(s)"),
         ("lint", "WARN lint", "Lint findings"),
-        ("analyze", "PASS analyze", "Analysis details"),
+        ("analyze", "PASS analyze", "Program overview"),
         ("inspect", "PASS inspect", "Program structure"),
     ] {
         let output = run(&[
@@ -187,6 +187,34 @@ fn terminal_renderer_uses_command_specific_hierarchy() {
     let _ = std::fs::remove_dir_all(path.parent().unwrap());
 }
 
+#[test]
+fn analyze_real_project_report_is_bounded_and_ranked() {
+    let path = temp_file(
+        "pixelart.txt",
+        &corpus_workshop("real-world/overpy-pixelart"),
+    );
+    let output = run(&[
+        "analyze",
+        path.to_str().unwrap(),
+        "--renderer",
+        "terminal",
+        "--color",
+        "never",
+    ]);
+    assert!(output.status.success());
+    let stdout = String::from_utf8_lossy(&output.stdout);
+    assert!(stdout.contains("Program overview"));
+    assert!(stdout.contains("Control-flow summary"));
+    assert!(stdout.contains("Top rules (heuristic ranking"));
+    assert!(stdout.contains("State and coupling"));
+    assert!(stdout.contains("[static]"));
+    assert!(
+        stdout.lines().count() <= 40,
+        "report is not bounded:\n{stdout}"
+    );
+    let _ = std::fs::remove_dir_all(path.parent().unwrap());
+}
+
 #[test]
 fn check_over_malformed_input_exits_one_with_structured_diagnostics() {
     // Enough locale evidence to pass detection, then a syntax error.
@@ -586,6 +614,14 @@ fn version_and_help_are_documented_contract_surfaces() {
     for command in ["compile", "convert", "check", "analyze", "lint", "inspect"] {
         assert!(help.contains(command), "help documents {command}");
     }
+    assert!(
+        help.contains("semantic hotspots"),
+        "help distinguishes analyze"
+    );
+    assert!(
+        help.contains("exhaustive structural"),
+        "help distinguishes inspect"
+    );
     for option in [
         "--kind",
         "--target",
diff --git a/docs/cli.md b/docs/cli.md
index d64cae9..ef69fca 100644
--- a/docs/cli.md
+++ b/docs/cli.md
@@ -135,9 +135,9 @@ result.
 | `wright compile [INPUT]` | Parse, lower, validate, emit Workshop text | the emitted artifact (or nothing with `-o`) |
 | `wright convert [INPUT] --target opy\|ostw` | Reconstruct validated Workshop input as canonical OPY or OSTW source | the reconstructed source |
 | `wright check [INPUT]` | Parse, lower, validate, and report correctness diagnostics | verdict and validation diagnostics |
-| `wright analyze [INPUT]` | Report semantic structure, symbol usage, and CFG measurements | semantic facts and measurements |
+| `wright analyze [INPUT]` | Summarize project structure, ranked CFG hotspots, and cross-cutting state | bounded semantic report with static evidence labels |
 | `wright lint [INPUT]` | Parse, lower, lint; report findings | findings, rule metadata, and effective-configuration summary |
-| `wright inspect [INPUT]` | Parse, lower, inspect structure | rules, symbols, references summary |
+| `wright inspect [INPUT]` | Parse, lower, and inspect exhaustive semantic facts | rules, symbols, references summary |
 | `wright completion <SHELL>` | Generate static completion script for bash, zsh, fish, or powershell | the generated completion script |
 | `wright completion install [SHELL]` | Install generated completion into standard user-local directory | installation progress and guidance |
 | `wright update` | Self-update a standalone installation | update progress (text only) |
@@ -311,11 +311,14 @@ The three core workflows have separate contracts:
 * `lint` executes the configurable `LintRegistry` and returns stable rule IDs,
   severity, evidence class, boundedness where applicable, source spans, rule
   metadata, and effective configuration.
-* `analyze` returns semantic facts rather than lint findings. Its initial
-  `result.facts` report contains symbol usage (`reads`, `writes`, `calls`, and
-  referencing rule count) and per-rule CFG measurements (blocks, edges, loop
-  blocks, and wait blocks). These facts can inform future lint rules without
-  making analysis a view of the registry.
+* `analyze` returns semantic facts rather than lint findings. Human text output
+  is a bounded report with a program overview, aggregate CFG measurements,
+  ranked rule hotspots, and ranked cross-cutting variables. The displayed
+  facts are static; rankings are heuristics based on CFG size or usage
+  coupling. `analyze --format json` retains the complete `result.facts`
+  payload for agents and embedding, while `inspect` is the human-facing
+  exhaustive structural/semantic view. These facts can inform future lint
+  rules without making analysis a view of the registry.
 
 Analysis findings (`lint` and the tool/agent `getFindings`/`lint` responses)
 carry an `evidence` field classifying how strongly the finding is supported