diff --git a/py/api/inspection_service.py b/py/api/inspection_service.py
index 0b15bd2..83ce648 100644
--- a/py/api/inspection_service.py
+++ b/py/api/inspection_service.py
@@ -9,6 +9,7 @@ try:
         MetadataReadError,
         inspect_prompt_evidence,
         read_inspection_file,
+        unreadable_evidence,
     )
     from ...utils.video_metadata import is_video_path
 except ImportError:
@@ -16,6 +17,7 @@ except ImportError:
         MetadataReadError,
         inspect_prompt_evidence,
         read_inspection_file,
+        unreadable_evidence,
     )
     from utils.video_metadata import is_video_path
 
@@ -73,19 +75,7 @@ def inspect_current_file(
     if any(getattr(before, key) != getattr(after, key) for key in fields):
         raise CurrentFileChanged
     if inspection is None:
-        evidence = {
-            "prompts": {
-                role: {"status": "unresolved", "text": None, "source": None}
-                for role in ("positive", "negative")
-            },
-            "resolution": {
-                "state": "unresolved",
-                "sampler_id": None,
-                "reason": "Media metadata could not be read",
-                "reason_code": None,
-                "node": None,
-            },
-        }
+        evidence = unreadable_evidence()
     else:
         role_saved = (
             None if saved and saved.get("association_status") == "ambiguous" else saved
diff --git a/tests/browser/prompt-inspector.spec.mjs b/tests/browser/prompt-inspector.spec.mjs
index e509be9..c211c8b 100644
--- a/tests/browser/prompt-inspector.spec.mjs
+++ b/tests/browser/prompt-inspector.spec.mjs
@@ -232,6 +232,17 @@ test('unverified saved associations explain uncertainty and cannot be sent',asyn
     await expect(details(page).getByRole('button',{name:/^Send /})).toHaveCount(0);
 });
 
+test('unresolved roles name the blocking node as text and offer no prompt actions',async({page})=>{
+    const line='Prompt uses a node the inspector does not support (<img src=x onerror="window.injected=1">).';
+    await page.route('**/images/inspect/uncertain.png?root=0*',route=>route.fulfill({json:{success:true,root_scope:new URL(route.request().url()).searchParams.get('root_scope'),image:{root_index:0,path:'uncertain.png',filename:'uncertain.png'},prompts:{
+        positive:{status:'unresolved',text:null,source:null,message:line},negative:{status:'unresolved',text:null,source:null,message:line},
+    }}}));
+    await page.goto(workspace);await select(page,'0:uncertain.png');
+    for (const name of ['Positive','Negative']) await expect(role(details(page),name)).toContainText(line);
+    await expect(details(page).locator('img[onerror]')).toHaveCount(0);
+    expect(await page.evaluate(()=>window.injected)).toBeUndefined();
+    await expect(details(page).getByRole('button',{name:/^(Send|Copy) /})).toHaveCount(0);
+});
 
 test('prompt role headings contain compact real actions and honest source evidence',async({page})=>{
     await page.goto(workspace);await select(page,'0:shared/same.png');
diff --git a/tests/js/prompt-blocks.test.js b/tests/js/prompt-blocks.test.js
index dea91bd..b32c6be 100644
--- a/tests/js/prompt-blocks.test.js
+++ b/tests/js/prompt-blocks.test.js
@@ -81,3 +81,52 @@ test('a zeroed conditioning negative shows as an explicit empty negative with it
         assert.deepEqual(sent,[{positive:'positive',negative:''}]);
     }finally{scope.dispose();f.restore();}
 });
+
+const unresolved=message=>({status:'unresolved',text:null,source:null,...(message===undefined?{}:{message})});
+const statusLines=(f,container)=>f.walk(container,node=>node.tagName==='p'&&node.className==='muted').map(node=>node.textContent);
+test('unsupported node status names the class',()=>{
+    const f=dom(),scope=createScope(),container=new f.Element('div');
+    const line='Prompt uses a node the inspector does not support (DualCFGGuider).';
+    try{
+        module.mountPromptBlocks(container,{parentScope:scope,sender:{open:()=>{}},prompts:{positive:unresolved(line),negative:unresolved(line)}});
+        assert.deepEqual(statusLines(f,container),[line,line]);
+        assert.equal(f.walk(container,node=>node.dataset.sendRole!==undefined||node.dataset.copyRole!==undefined).length,0,'no action implies a prompt exists');
+    }finally{scope.dispose();f.restore();}
+});
+test('runtime and composite statuses',()=>{
+    const f=dom(),scope=createScope(),container=new f.Element('div');
+    const runtime='Prompt is chosen at run time (PromptSearchList).',composite='Prompt combines several texts (ConditioningCombine).';
+    try{
+        const view=module.mountPromptBlocks(container,{parentScope:scope,sender:{open:()=>{}},prompts:{positive:unresolved(runtime),negative:unresolved(runtime)}});
+        assert.deepEqual(statusLines(f,container),[runtime,runtime]);
+        view.update({positive:unresolved(composite),negative:{status:'resolved',text:'kept',source:'execution'}});
+        assert.deepEqual(statusLines(f,container),[composite]);
+        assert.equal(f.walk(container,node=>node.tagName==='pre'&&node.textContent==='kept').length,1);
+    }finally{scope.dispose();f.restore();}
+});
+test('limit status',()=>{
+    const f=dom(),scope=createScope(),container=new f.Element('div');
+    const line='Prompt exceeds inspection limits.';
+    try{
+        module.mountPromptBlocks(container,{parentScope:scope,sender:{open:()=>{}},prompts:{positive:unresolved(line),negative:unresolved(line)}});
+        assert.deepEqual(statusLines(f,container),[line,line]);
+    }finally{scope.dispose();f.restore();}
+});
+test('roles without a usable line keep the status copy, and a candidate is never shown as a prompt',()=>{
+    const f=dom(),scope=createScope(),container=new f.Element('div');
+    try{
+        module.mountPromptBlocks(container,{parentScope:scope,sender:{open:()=>{}},prompts:{positive:unresolved(''),negative:{status:'ambiguous',text:'candidate text',source:null}}});
+        assert.deepEqual(statusLines(f,container),['Prompt could not be resolved.','Multiple possible prompts. Source could not be determined.']);
+        assert.equal(f.walk(container,node=>node.textContent==='candidate text').length,0);
+    }finally{scope.dispose();f.restore();}
+});
+test('class names in status lines render as text, never markup',()=>{
+    const f=dom(),scope=createScope(),container=new f.Element('div');
+    const line='Prompt uses a node the inspector does not support (<img src=x onerror=alert(1)>).';
+    try{
+        module.mountPromptBlocks(container,{parentScope:scope,sender:{open:()=>{}},prompts:{positive:unresolved(line),negative:unresolved(line)}});
+        assert.deepEqual(statusLines(f,container),[line,line]);
+        assert.equal(f.walk(container,node=>node.tagName==='img').length,0);
+        assert.equal(f.walk(container,node=>Object.hasOwn(node,'innerHTML')).length,0);
+    }finally{scope.dispose();f.restore();}
+});
diff --git a/tests/js/workspace-inspection-state.test.js b/tests/js/workspace-inspection-state.test.js
index f327e23..fcb8e4d 100644
--- a/tests/js/workspace-inspection-state.test.js
+++ b/tests/js/workspace-inspection-state.test.js
@@ -225,6 +225,48 @@ test('record accepts closed reason codes and rejects contradictory ones', () =>
     }
 });
 
+test('resolution validator accepts closed reason codes and a bounded node record, rejects others', () => {
+    const toRecord = contract('toInspectionRecord');
+    for (const reason_code of ['unsupported-node', 'runtime-value', 'composite']) {
+        for (const node of [{ id: '7', class_type: 'N'.repeat(200) }, { id: '7'.repeat(200), class_type: null }, null]) {
+            const data = envelope(); Object.assign(data.resolution, { state: 'unresolved', reason_code, node });
+            assert.doesNotThrow(() => toRecord(data, image(), extractGenerationFacts), JSON.stringify({ reason_code, node }));
+        }
+    }
+    for (const change of [
+        { reason_code: 'unsupported-node', node: { id: '7', class_type: 'N'.repeat(201) } },
+        { reason_code: 'unsupported-node', node: { id: '7'.repeat(201), class_type: 'X' } },
+        { reason_code: 'unsupported-node', node: { id: '', class_type: 'X' } },
+        { reason_code: 'unsupported-node', node: { id: '7', class_type: 'X', extra: true } },
+        { reason_code: 'unsupported-node', node: [] },
+        { reason_code: 'malformed', node: { id: '7', class_type: 'X' } },
+        { reason_code: 'work-limit', node: { id: '7', class_type: 'X' } },
+        { reason_code: 'no-sampler', node: { id: '7', class_type: 'X' } },
+        { reason_code: 'Unsupported-Node', node: null },
+        { reason_code: '', node: null },
+    ]) {
+        const data = envelope(); Object.assign(data.resolution, { state: 'unresolved', ...change });
+        assert.throws(() => toRecord(data, image(), extractGenerationFacts), TypeError, JSON.stringify(change));
+    }
+});
+
+test('role outcome lines are bounded text on non-resolved roles only', () => {
+    const toRecord = contract('toInspectionRecord');
+    const line = 'Prompt uses a node the inspector does not support (<img src=x onerror=alert(1)>).';
+    const data = envelope(); data.prompts.negative = { status: 'unresolved', text: null, source: null, message: line };
+    assert.equal(toRecord(data, image(), extractGenerationFacts).prompts.negative.message, line);
+    for (const [role, record] of [
+        ['positive', { status: 'resolved', text: 'exact', source: 'execution', message: 'Hidden reason.' }],
+        ['negative', { status: 'unresolved', text: null, source: null, message: '' }],
+        ['negative', { status: 'unresolved', text: null, source: null, message: 7 }],
+        ['negative', { status: 'unresolved', text: null, source: null, message: null }],
+        ['negative', { status: 'unresolved', text: null, source: null, message: 'x'.repeat(513) }],
+    ]) {
+        const bad = envelope(); bad.prompts[role] = record;
+        assert.throws(() => toRecord(bad, image(), extractGenerationFacts), TypeError, JSON.stringify(record));
+    }
+});
+
 test('record rejects root path media and invalid versioned fields with retryable TypeError', () => {
     const toRecord = contract('toInspectionRecord');
     for (const mutate of [
diff --git a/tests/test_api_prompt_inspector.py b/tests/test_api_prompt_inspector.py
index f36c962..cc41e1b 100644
--- a/tests/test_api_prompt_inspector.py
+++ b/tests/test_api_prompt_inspector.py
@@ -244,10 +244,10 @@ class TestPromptInspector(ImageAPITestCase):
 
     async def test_corrupt_malformed_and_no_metadata_do_not_invent_text(self):
         path = self.output_dir / "image.png"
-        for content, expected in (
-            (b"broken", "unresolved"),
-            (None, "absent"),
-            ("bad-json", "unresolved"),
+        for content, expected, message in (
+            (b"broken", "unresolved", "Metadata in this file could not be read."),
+            (None, "absent", None),
+            ("bad-json", "unresolved", "Prompt metadata in this file is malformed."),
         ):
             if isinstance(content, bytes):
                 path.write_bytes(content)
@@ -256,6 +256,7 @@ class TestPromptInspector(ImageAPITestCase):
             _, result = await self.inspect()
             self.assertEqual(result["prompts"]["positive"]["status"], expected)
             self.assertIsNone(result["prompts"]["positive"]["text"])
+            self.assertEqual(result["prompts"]["negative"].get("message"), message)
 
     async def test_compressed_international_and_post_pixel_metadata(self):
         import struct
@@ -790,6 +791,9 @@ class TestPromptInspector(ImageAPITestCase):
             result["resolution"]["reason"],
             "No supported sampler pair found (saw FutureSampler)",
         )
+        self.assertEqual(
+            result["prompts"]["positive"]["message"], "No supported sampler was found."
+        )
 
     async def test_inspect_route_exposes_reason_code_and_blocking_node(self):
         # f06a's encoder is supported since Task 2.4; an unknown encoder is not.
@@ -803,6 +807,14 @@ class TestPromptInspector(ImageAPITestCase):
         self.assertEqual(
             resolution["node"], {"id": "6", "class_type": "FutureTextEncode"}
         )
+        self.assertEqual(
+            {role: side.get("message") for role, side in result["prompts"].items()},
+            dict.fromkeys(
+                ("positive", "negative"),
+                "Prompt uses a node the inspector does not support "
+                "(FutureTextEncode).",
+            ),
+        )
         png(self.output_dir / "image.png", prompt="{broken")
         _, result = await self.inspect()
         self.assertEqual(result["resolution"]["reason_code"], "malformed")
diff --git a/tests/test_inspection_outcomes.py b/tests/test_inspection_outcomes.py
new file mode 100644
index 0000000..2420ea9
--- /dev/null
+++ b/tests/test_inspection_outcomes.py
@@ -0,0 +1,355 @@
+"""Non-resolved prompt outcomes carry one short, truthful line per role."""
+
+import json
+import unittest
+from dataclasses import replace
+from unittest import mock
+
+from utils.inspection_outcomes import (
+    CLASS_NAME_LENGTH,
+    REASON_CODES,
+    outcome_message,
+    project_resolution,
+)
+from utils.inspection_response import encode_inspection_response
+from utils.prompt_inspector import inspect_prompt_evidence
+from utils.prompt_pair_resolver import (
+    EMPTY_POSITIVE_REASON,
+    INSPECTION_PROFILE,
+    MULTISTAGE_UNATTRIBUTED_REASON,
+    SEMANTICS_UNAVAILABLE_REASON,
+)
+
+LIMITS = "Prompt exceeds inspection limits."
+
+
+def _encode(text: str) -> dict:
+    return {"class_type": "CLIPTextEncode", "inputs": {"text": text, "clip": ["9", 0]}}
+
+
+def _latent() -> dict:
+    return {
+        "class_type": "EmptyLatentImage",
+        "inputs": {"width": 8, "height": 8, "batch_size": 1},
+    }
+
+
+def _ksampler(positive: str = "7") -> dict:
+    return {
+        "class_type": "KSampler",
+        "inputs": {
+            "positive": [positive, 0],
+            "negative": ["6", 0],
+            "latent_image": ["4", 0],
+            "model": ["9", 0],
+            "seed": 1,
+            "steps": 2,
+            "cfg": 1.0,
+            "sampler_name": "euler",
+            "scheduler": "normal",
+            "denoise": 1.0,
+        },
+    }
+
+
+def _dual_cfg() -> dict:
+    return {
+        "1": _encode("positive"),
+        "6": _encode("negative"),
+        "2": {
+            "class_type": "DualCFGGuider",
+            "inputs": {
+                "cond1": ["1", 0],
+                "cond2": ["1", 0],
+                "negative": ["6", 0],
+                "model": ["9", 0],
+            },
+        },
+        "3": {
+            "class_type": "SamplerCustomAdvanced",
+            "inputs": {
+                "guider": ["2", 0],
+                "noise": ["9", 0],
+                "sampler": ["9", 0],
+                "sigmas": ["9", 0],
+                "latent_image": ["4", 0],
+            },
+        },
+        "4": _latent(),
+    }
+
+
+def _runtime() -> dict:
+    return {
+        "1": {"class_type": "PromptSearchList", "inputs": {"search": "cat"}},
+        "7": {
+            "class_type": "CLIPTextEncode",
+            "inputs": {"text": ["1", 0], "clip": ["9", 0]},
+        },
+        "6": _encode("negative"),
+        "3": _ksampler(),
+        "4": _latent(),
+    }
+
+
+def _composite() -> dict:
+    graph = _runtime()
+    graph["7"] = {
+        "class_type": "ConditioningCombine",
+        "inputs": {"conditioning_1": ["6", 0], "conditioning_2": ["8", 0]},
+    }
+    graph["8"] = _encode("other")
+    return graph
+
+
+def _messages(evidence: dict) -> dict:
+    return {role: side.get("message") for role, side in evidence["prompts"].items()}
+
+
+def _statuses(evidence: dict) -> dict:
+    return {role: side["status"] for role, side in evidence["prompts"].items()}
+
+
+class TestOutcomeMessages(unittest.TestCase):
+    def assert_unresolved_with(self, evidence: dict, message: str) -> None:
+        self.assertEqual(
+            _statuses(evidence), {"positive": "unresolved", "negative": "unresolved"}
+        )
+        self.assertEqual(
+            _messages(evidence), {"positive": message, "negative": message}
+        )
+
+    def test_unsupported_node_status_names_the_class(self):
+        evidence = inspect_prompt_evidence({"prompt": _dual_cfg()})
+        self.assertEqual(evidence["resolution"]["reason_code"], "unsupported-node")
+        self.assert_unresolved_with(
+            evidence,
+            "Prompt uses a node the inspector does not support (DualCFGGuider).",
+        )
+
+    def test_runtime_and_composite_statuses(self):
+        self.assert_unresolved_with(
+            inspect_prompt_evidence({"prompt": _runtime()}),
+            "Prompt is chosen at run time (PromptSearchList).",
+        )
+        self.assert_unresolved_with(
+            inspect_prompt_evidence({"prompt": _composite()}),
+            "Prompt combines several texts (ConditioningCombine).",
+        )
+
+    def test_limit_status(self):
+        tight = replace(INSPECTION_PROFILE, per_candidate_visits=1, total_visits=1)
+        evidence = inspect_prompt_evidence({"prompt": _runtime()}, profile=tight)
+        self.assertEqual(evidence["resolution"]["reason_code"], "work-limit")
+        self.assert_unresolved_with(evidence, LIMITS)
+        for code in ("work-limit", "depth-limit", "text-limit", "graph-ceiling"):
+            self.assertEqual(
+                outcome_message({"state": "unresolved", "reason_code": code}), LIMITS
+            )
+
+    def test_every_closed_code_and_known_uncoded_reason_has_a_line(self):
+        for code in REASON_CODES:
+            message = outcome_message({"state": "unresolved", "reason_code": code})
+            self.assertIsInstance(message, str, code)
+            self.assertTrue(message.endswith("."), code)
+            self.assertNotIn("could not be resolved", message)
+        for reason, expected in (
+            (EMPTY_POSITIVE_REASON, "Positive prompt is empty."),
+            (
+                SEMANTICS_UNAVAILABLE_REASON,
+                "The inspector's node table is unavailable.",
+            ),
+            (
+                MULTISTAGE_UNATTRIBUTED_REASON,
+                "Several sampler stages were found; their prompts were not compared.",
+            ),
+            (
+                "Media metadata could not be read",
+                "Metadata in this file could not be read.",
+            ),
+        ):
+            resolution = {"state": "unresolved", "reason": reason, "reason_code": None}
+            self.assertEqual(outcome_message(resolution), expected)
+
+    def test_named_codes_without_a_class_omit_the_name(self):
+        for code, line in (
+            ("unsupported-node", "Prompt uses a node the inspector does not support."),
+            ("runtime-value", "Prompt is chosen at run time."),
+            ("composite", "Prompt combines several texts."),
+        ):
+            for node in (None, {"id": "7", "class_type": None}):
+                resolution = {"state": "unresolved", "reason_code": code, "node": node}
+                self.assertEqual(outcome_message(resolution), line)
+
+    def test_class_name_is_bounded(self):
+        resolution = {
+            "state": "unresolved",
+            "reason_code": "unsupported-node",
+            "node": {"id": "7", "class_type": "N" * 500},
+        }
+        self.assertEqual(
+            outcome_message(resolution),
+            "Prompt uses a node the inspector does not support "
+            f"({'N' * CLASS_NAME_LENGTH}).",
+        )
+        self.assertEqual(CLASS_NAME_LENGTH, 200)
+
+    def test_resolved_ambiguous_absent_and_unknown_outcomes_carry_no_line(self):
+        for resolution in (
+            {"state": "resolved", "reason_code": None},
+            {"state": "ambiguous", "reason_code": "multistage-differs"},
+            {"state": "ambiguous", "reason_code": None, "reason": "Multiple"},
+            {"state": "absent", "reason_code": None, "reason": "No parsed"},
+            {"state": "unresolved", "reason_code": "guess"},
+            {"state": "unresolved", "reason_code": None, "reason": "Anything else"},
+            None,
+        ):
+            self.assertIsNone(outcome_message(resolution), resolution)
+
+    def test_inspector_lines_for_no_sampler_malformed_and_empty_positive(self):
+        evidence = inspect_prompt_evidence({"prompt": {"1": _encode("x")}})
+        self.assert_unresolved_with(evidence, "No supported sampler was found.")
+        evidence = inspect_prompt_evidence({"prompt": "not-json"})
+        self.assert_unresolved_with(
+            evidence, "Prompt metadata in this file is malformed."
+        )
+        graph = _runtime()
+        graph["7"] = _encode("   ")
+        evidence = inspect_prompt_evidence({"prompt": graph})
+        self.assertEqual(evidence["resolution"]["reason"], EMPTY_POSITIVE_REASON)
+        self.assert_unresolved_with(evidence, "Positive prompt is empty.")
+
+    def test_a_resolved_pair_with_an_empty_positive_explains_only_that_role(self):
+        side = {"text": " ", "provenance": {"text_source": "graph-inferred"}}
+        negative = {"text": "kept", "provenance": {"text_source": "graph-inferred"}}
+        resolved = {
+            "state": "resolved",
+            "positive": side,
+            "negative": negative,
+            "sampler_id": "3",
+            "reason": "",
+            "reason_code": None,
+            "node": None,
+        }
+        with mock.patch(
+            "utils.prompt_inspector.resolve_prompt_pair", return_value=resolved
+        ):
+            evidence = inspect_prompt_evidence({"prompt": {}})
+        self.assertEqual(
+            evidence["prompts"]["positive"],
+            {
+                "status": "unresolved",
+                "text": None,
+                "source": None,
+                "message": "Positive prompt is empty.",
+            },
+        )
+        self.assertEqual(evidence["prompts"]["negative"]["text"], "kept")
+        self.assertNotIn("message", evidence["prompts"]["negative"])
+
+    def test_candidates_and_absence_keep_their_status_lines(self):
+        graph = _runtime()
+        graph["7"] = _encode("positive")
+        graph["5"] = _ksampler()
+        evidence = inspect_prompt_evidence({"prompt": graph})
+        self.assertEqual(evidence["resolution"]["state"], "ambiguous")
+        self.assertEqual(_messages(evidence), {"positive": None, "negative": None})
+        evidence = inspect_prompt_evidence({})
+        self.assertEqual(_messages(evidence), {"positive": None, "negative": None})
+        graph = _runtime()
+        graph["7"] = _encode("positive")
+        evidence = inspect_prompt_evidence({"prompt": graph})
+        self.assertEqual(evidence["resolution"]["state"], "resolved")
+        self.assertEqual(_messages(evidence), {"positive": None, "negative": None})
+
+
+class TestResolutionProjection(unittest.TestCase):
+    def test_closed_codes_and_named_nodes_pass_unchanged(self):
+        for code in REASON_CODES:
+            resolution = {
+                "state": "unresolved",
+                "sampler_id": "3",
+                "reason": "r",
+                "reason_code": code,
+                "node": None,
+            }
+            self.assertEqual(project_resolution(resolution), resolution)
+        named = {
+            "state": "unresolved",
+            "sampler_id": None,
+            "reason": "r",
+            "reason_code": "composite",
+            "node": {"id": "7", "class_type": None},
+        }
+        self.assertEqual(project_resolution(named), named)
+
+    def test_unknown_codes_and_unbounded_nodes_fail_closed(self):
+        base = {"state": "unresolved", "sampler_id": None, "reason": "r"}
+        for code, node in (
+            ("guess", None),
+            (1, None),
+            ("malformed", {"id": "7", "class_type": "X"}),
+            (None, {"id": "7", "class_type": "X"}),
+            ("unsupported-node", {"id": 7, "class_type": "X"}),
+            ("unsupported-node", {"id": "", "class_type": "X"}),
+            ("unsupported-node", {"id": "7"}),
+            ("unsupported-node", {"id": "7", "class_type": 7}),
+            ("unsupported-node", {"id": "7", "class_type": "X", "extra": 1}),
+            ("unsupported-node", "X"),
+        ):
+            projected = project_resolution({**base, "reason_code": code, "node": node})
+            self.assertIsNone(projected["node"], (code, node))
+            if code not in REASON_CODES:
+                self.assertIsNone(projected["reason_code"], code)
+        for state in ("resolved", "absent"):
+            projected = project_resolution(
+                {**base, "state": state, "reason_code": "work-limit", "node": None}
+            )
+            self.assertIsNone(projected["reason_code"])
+        long_name = project_resolution(
+            {
+                **base,
+                "reason_code": "runtime-value",
+                "node": {"id": "7" * 300, "class_type": "N" * 300},
+            }
+        )["node"]
+        self.assertEqual(
+            long_name, {"id": "7" * CLASS_NAME_LENGTH, "class_type": "N" * 200}
+        )
+        unnamed = project_resolution(
+            {
+                **base,
+                "reason_code": "composite",
+                "node": {"id": "7", "class_type": ""},
+            }
+        )["node"]
+        self.assertEqual(unnamed, {"id": "7", "class_type": None})
+
+    def test_wire_carries_role_lines_and_projected_resolution(self):
+        evidence = inspect_prompt_evidence({"prompt": _dual_cfg()})
+        resolution = dict(evidence["resolution"], reason_code="guess")
+        body = json.loads(
+            encode_inspection_response(
+                image={"root_index": 0, "path": "a.png", "filename": "a.png"},
+                root_scope="a" * 64,
+                file={
+                    "observation": "current-file",
+                    "source_label": "output",
+                    "media_type": "image",
+                    "extension": ".png",
+                    "size": 1,
+                    "modified_time": 1,
+                },
+                prompts=evidence["prompts"],
+                resolution=resolution,
+                inspection=None,
+                library={"status": "none", "tags": []},
+            )
+        )
+        self.assertEqual(body["prompts"], evidence["prompts"])
+        self.assertIsNone(body["resolution"]["reason_code"])
+        self.assertIsNone(body["resolution"]["node"])
+
+
+if __name__ == "__main__":
+    unittest.main()
diff --git a/utils/inspection_outcomes.py b/utils/inspection_outcomes.py
new file mode 100644
index 0000000..27eadf6
--- /dev/null
+++ b/utils/inspection_outcomes.py
@@ -0,0 +1,109 @@
+"""User-facing lines for non-resolved prompt outcomes, from the closed vocabulary.
+
+The inspector shows one line per unresolved role in place of the generic
+"Prompt could not be resolved." Ambiguous and absent roles keep their status
+copy, so a candidate is never presented as a confirmed prompt. The wire
+projection accepts exactly the closed ``reason_code`` vocabulary and names a
+blocking node only beside a code that names one.
+"""
+
+from typing import Any, Optional
+
+from .prompt_pair_resolver import (
+    EMPTY_POSITIVE_REASON,
+    MULTISTAGE_UNATTRIBUTED_REASON,
+    SEMANTICS_UNAVAILABLE_REASON,
+)
+
+REASON_CODES = frozenset(
+    {
+        "work-limit",
+        "depth-limit",
+        "text-limit",
+        "graph-ceiling",
+        "unsupported-node",
+        "runtime-value",
+        "composite",
+        "no-sampler",
+        "malformed",
+        "multistage-differs",
+    }
+)
+LIMIT_CODES = frozenset({"work-limit", "depth-limit", "text-limit", "graph-ceiling"})
+# Codes whose outcome names the blocking node: ``node: {id, class_type}``.
+NAMED_CODES = frozenset({"unsupported-node", "runtime-value", "composite"})
+# Bound for a node ID or class name on the wire and in a line.
+CLASS_NAME_LENGTH = 200
+UNREADABLE_METADATA_REASON = "Media metadata could not be read"
+_NODE_KEYS = frozenset({"id", "class_type"})
+
+_NAMED_LINES = {
+    "unsupported-node": "Prompt uses a node the inspector does not support",
+    "runtime-value": "Prompt is chosen at run time",
+    "composite": "Prompt combines several texts",
+}
+_CODED_LINES = {
+    **dict.fromkeys(LIMIT_CODES, "Prompt exceeds inspection limits."),
+    "no-sampler": "No supported sampler was found.",
+    "malformed": "Prompt metadata in this file is malformed.",
+    "multistage-differs": "Sampler stages have different prompts.",
+}
+# Uncoded outcomes, keyed by the resolver's exact reason.
+_UNCODED_LINES = {
+    EMPTY_POSITIVE_REASON: "Positive prompt is empty.",
+    MULTISTAGE_UNATTRIBUTED_REASON: (
+        "Several sampler stages were found; their prompts were not compared."
+    ),
+    SEMANTICS_UNAVAILABLE_REASON: "The inspector's node table is unavailable.",
+    UNREADABLE_METADATA_REASON: "Metadata in this file could not be read.",
+}
+
+
+def _text(value: Any) -> Optional[str]:
+    return value if isinstance(value, str) else None
+
+
+def _class_name(node: Any) -> Optional[str]:
+    if not isinstance(node, dict):
+        return None
+    name = node.get("class_type")
+    return name[:CLASS_NAME_LENGTH] if isinstance(name, str) and name else None
+
+
+def outcome_message(resolution: Any) -> Optional[str]:
+    """One line for an unresolved outcome, or ``None`` to keep the status copy."""
+    if not isinstance(resolution, dict) or resolution.get("state") != "unresolved":
+        return None
+    code = _text(resolution.get("reason_code"))
+    if code in _NAMED_LINES:
+        name = _class_name(resolution.get("node"))
+        return f"{_NAMED_LINES[code]} ({name})." if name else f"{_NAMED_LINES[code]}."
+    if code is None:
+        return _UNCODED_LINES.get(_text(resolution.get("reason")))
+    return _CODED_LINES.get(code)
+
+
+def _node(code: Optional[str], node: Any) -> Optional[dict]:
+    if code not in NAMED_CODES or not isinstance(node, dict) or set(node) != _NODE_KEYS:
+        return None
+    node_id, name = node["id"], node["class_type"]
+    if not isinstance(node_id, str) or not node_id:
+        return None
+    if name is not None and not isinstance(name, str):
+        return None
+    return {
+        "id": node_id[:CLASS_NAME_LENGTH],
+        "class_type": name[:CLASS_NAME_LENGTH] if name else None,
+    }
+
+
+def project_resolution(resolution: dict) -> dict:
+    """Fail closed: an unknown or contradictory code and its node are dropped."""
+    code = _text(resolution.get("reason_code"))
+    if code not in REASON_CODES or resolution.get("state") in ("resolved", "absent"):
+        code = None
+    return {
+        **resolution,
+        "reason_code": code,
+        "node": _node(code, resolution.get("node")),
+    }
diff --git a/utils/inspection_response.py b/utils/inspection_response.py
index b8a4669..357e97b 100644
--- a/utils/inspection_response.py
+++ b/utils/inspection_response.py
@@ -5,6 +5,8 @@ import math
 from decimal import Decimal, InvalidOperation
 from typing import TYPE_CHECKING, Literal
 
+from .inspection_outcomes import project_resolution
+
 if TYPE_CHECKING:
     from .prompt_inspector import InspectionRead
 
@@ -329,7 +331,7 @@ def _encode_inspection(
         **identity,
         "file": _file_projection(file, inspection),
         "prompts": prompts,
-        "resolution": resolution,
+        "resolution": project_resolution(resolution),
         "metadata": metadata,
         "library": _library_projection(library),
     }
diff --git a/utils/prompt_inspector.py b/utils/prompt_inspector.py
index f38aa9a..5646b98 100644
--- a/utils/prompt_inspector.py
+++ b/utils/prompt_inspector.py
@@ -13,8 +13,10 @@ from typing import Any
 
 from PIL import Image
 
+from .inspection_outcomes import UNREADABLE_METADATA_REASON, outcome_message
 from .pair_execution import EXECUTION_KEY
 from .prompt_pair_resolver import (
+    EMPTY_POSITIVE_REASON,
     INSPECTION_PROFILE,
     ResolverProfile,
     resolve_prompt_pair,
@@ -400,6 +402,39 @@ def _side(status: str, text: str | None = None, source: str | None = None) -> di
     return {"status": status, "text": text, "source": source}
 
 
+_EMPTY_POSITIVE = outcome_message(
+    {"state": "unresolved", "reason": EMPTY_POSITIVE_REASON}
+)
+
+
+def _explained(prompts: dict, resolution: dict) -> dict:
+    """Each unresolved role carries the outcome's line; other roles are kept."""
+    message = outcome_message(resolution)
+    if message is None:
+        return prompts
+    return {
+        role: (
+            {**side, "message": message}
+            if side["status"] == "unresolved" and "message" not in side
+            else side
+        )
+        for role, side in prompts.items()
+    }
+
+
+def unreadable_evidence() -> dict:
+    """Both roles when the file's metadata could not be read at all."""
+    resolution = {
+        "state": "unresolved",
+        "sampler_id": None,
+        "reason": UNREADABLE_METADATA_REASON,
+        "reason_code": None,
+        "node": None,
+    }
+    prompts = {role: _side("unresolved") for role in ("positive", "negative")}
+    return {"prompts": _explained(prompts, resolution), "resolution": resolution}
+
+
 def inspect_prompt_evidence(
     metadata: dict,
     saved: dict | None = None,
@@ -447,21 +482,20 @@ def inspect_prompt_evidence(
             },
         }
     canonical = dict(metadata)
+    malformed = {
+        "prompts": _explained(
+            {role: _side("unresolved") for role in ("positive", "negative")},
+            resolution,
+        ),
+        "resolution": resolution,
+    }
     if "prompt" not in canonical and "parameters" not in canonical:
-        return {
-            "prompts": {role: _side("unresolved") for role in ("positive", "negative")},
-            "resolution": resolution,
-        }
+        return malformed
     if isinstance(canonical.get("prompt"), str):
         try:
             canonical["prompt"] = json.loads(canonical["prompt"])
         except (ValueError, RecursionError):
-            return {
-                "prompts": {
-                    role: _side("unresolved") for role in ("positive", "negative")
-                },
-                "resolution": resolution,
-            }
+            return malformed
     resolved = resolve_prompt_pair(canonical, profile=profile)
     state = resolved["state"]
     result = {}
@@ -475,18 +509,19 @@ def inspect_prompt_evidence(
             result[role] = _side(
                 "resolved", side["text"], side["provenance"]["text_source"]
             )
+        elif state == "resolved" and isinstance(side, dict):
+            # A resolved pair withholds only an empty positive.
+            result[role] = {**_side("unresolved"), "message": _EMPTY_POSITIVE}
         else:
             result[role] = _side(state if state != "resolved" else "unresolved")
-    return {
-        "prompts": result,
-        "resolution": {
-            "state": state,
-            "sampler_id": resolved.get("sampler_id"),
-            "reason": resolved.get("reason", ""),
-            "reason_code": resolved.get("reason_code"),
-            "node": resolved.get("node"),
-        },
+    resolution = {
+        "state": state,
+        "sampler_id": resolved.get("sampler_id"),
+        "reason": resolved.get("reason", ""),
+        "reason_code": resolved.get("reason_code"),
+        "node": resolved.get("node"),
     }
+    return {"prompts": _explained(result, resolution), "resolution": resolution}
 
 
 def inspect_prompt_roles(metadata: dict, saved: dict | None = None) -> dict:
diff --git a/utils/prompt_pair_resolver.py b/utils/prompt_pair_resolver.py
index b6bef3e..f343602 100644
--- a/utils/prompt_pair_resolver.py
+++ b/utils/prompt_pair_resolver.py
@@ -151,6 +151,9 @@ _ZEROED_SOURCE = "zeroed-conditioning"
 _MAX_BRANCH_CANDIDATES = 32
 # Statically undecidable text (a search, a list output): never pick an item.
 _RUNTIME_REASON = "Text is decided at run time"
+# Uncoded outcomes the inspector explains by their exact reason.
+EMPTY_POSITIVE_REASON = "Positive prompt is empty"
+MULTISTAGE_UNATTRIBUTED_REASON = "Multistage sampler ancestry is ambiguous"
 _INCOMPATIBLE_SLOT = "Output slot has an incompatible type"
 # Trace outcomes that leave an input unverifiable against execution evidence.
 _UNVERIFIABLE_CODES = frozenset({"unsupported-node", "runtime-value"})
@@ -1189,7 +1192,7 @@ class _GraphResolver:
                 inputs.get(conditioning["positive"]), sampler_id, guider_id, depth
             )
             if not positive["text"].strip():
-                raise _Unresolved("Positive prompt is empty")
+                raise _Unresolved(EMPTY_POSITIVE_REASON)
             negative_name = conditioning["negative"]
             if negative_name is None or negative_name not in inputs:
                 return _result(
@@ -1227,7 +1230,7 @@ class _GraphResolver:
         name, inputs = self.node(node_id)
         if name in self.stage_classes:
             # ``multistage-differs`` is reserved for when stage texts are compared.
-            raise _Unresolved("Multistage sampler ancestry is ambiguous")
+            raise _Unresolved(MULTISTAGE_UNATTRIBUTED_REASON)
         if name in ("EmptyLatentImage", "EmptySD3LatentImage") and slot == 0:
             return
         if name == "Reroute" and slot == 0 and "value" in inputs:
@@ -1287,7 +1290,7 @@ class _GraphResolver:
             if name in self.stage_classes:
                 # ``multistage-differs`` is reserved for when stage texts are
                 # compared (PR3); until then no stage's prompt is chosen.
-                raise _Unresolved("Multistage sampler ancestry is ambiguous")
+                raise _Unresolved(MULTISTAGE_UNATTRIBUTED_REASON)
             if self.stage_shaped(name, inputs):
                 raise self.unsupported("Unsupported latent ancestry", node_id)
             self.enqueue(queue, queued, inputs, depth + 1)
@@ -1344,7 +1347,7 @@ def _parameters(value: Any, max_text_length: int) -> dict:
     positive, found, negative = body.partition(delimiter)
     positive = positive.strip()
     if not positive:
-        return _result("unresolved", reason="Positive prompt is empty")
+        return _result("unresolved", reason=EMPTY_POSITIVE_REASON)
     provenance = {
         "format": "a1111",
         "field": "parameters",
@@ -1424,7 +1427,7 @@ def resolve_prompt_pair(
     return _legacy_result(result) if profile.legacy else result
 
 
-_SEMANTICS_UNAVAILABLE = "Inspection node semantics are unavailable"
+SEMANTICS_UNAVAILABLE_REASON = "Inspection node semantics are unavailable"
 
 
 def _inspection_semantics() -> Optional[Any]:
@@ -1495,7 +1498,7 @@ def _resolve(
     if profile.semantics:
         semantics = _inspection_semantics()
         if semantics is None:
-            return _result("unresolved", reason=_SEMANTICS_UNAVAILABLE)
+            return _result("unresolved", reason=SEMANTICS_UNAVAILABLE_REASON)
     normalized = {str(node_id): node for node_id, node in graph.items()}
     if len(normalized) != len(graph):
         return _result(
diff --git a/web/js/workspace/inspection-state.mjs b/web/js/workspace/inspection-state.mjs
index 3997a1d..9b7fb6b 100644
--- a/web/js/workspace/inspection-state.mjs
+++ b/web/js/workspace/inspection-state.mjs
@@ -9,6 +9,10 @@ const FACT_FIELDS = ['seed', 'steps', 'cfg', 'sampler', 'scheduler', 'model'];
 const RECORD_FORMATS = ['original-text', 'reserialized-json', 'invalid', 'limited'];
 const TEXTLESS_FORMATS = ['invalid', 'limited'];
 const REASON_CODES = ['work-limit', 'depth-limit', 'text-limit', 'graph-ceiling', 'unsupported-node', 'runtime-value', 'composite', 'no-sampler', 'malformed', 'multistage-differs'];
+// Only these codes name the blocking node; node IDs and class names are bounded.
+const NAMED_CODES = ['unsupported-node', 'runtime-value', 'composite'];
+const NAME_LENGTH = 200, MESSAGE_LENGTH = 512;
+const boundedText = (value, limit) => typeof value === 'string' && value.length > 0 && value.length <= limit;
 const object = value => value !== null && typeof value === 'object' && !Array.isArray(value);
 const validScope = value => typeof value === 'string' && /^[a-f0-9]{64}$/.test(value) && value.length === 64;
 
@@ -77,7 +81,8 @@ function promptsFor(value) {
         if (record === undefined) continue;
         if (!object(record) || typeof record.status !== 'string' ||
             (record.text !== undefined && record.text !== null && typeof record.text !== 'string') ||
-            (record.status === 'resolved' && !Object.hasOwn(confirmed, role))) throw new TypeError('Invalid inspection role');
+            (record.status === 'resolved' && !Object.hasOwn(confirmed, role)) ||
+            (record.message !== undefined && (record.status === 'resolved' || !boundedText(record.message, MESSAGE_LENGTH)))) throw new TypeError('Invalid inspection role');
     }
     return value;
 }
@@ -141,7 +146,8 @@ function resolutionFor(value) {
     const code = value.reason_code ?? null, node = value.node ?? null;
     // Codes explain a non-resolved outcome; a blocking node is named only with its code.
     if ((code !== null && (!REASON_CODES.includes(code) || ['resolved', 'absent'].includes(value.state))) ||
-        (node !== null && (code === null || !object(node) || typeof node.id !== 'string' || (node.class_type !== null && typeof node.class_type !== 'string')))) throw new TypeError('Invalid inspection resolution reason');
+        (node !== null && (!NAMED_CODES.includes(code) || !object(node) || Object.keys(node).sort().join() !== 'class_type,id' ||
+            !boundedText(node.id, NAME_LENGTH) || (node.class_type !== null && !boundedText(node.class_type, NAME_LENGTH))))) throw new TypeError('Invalid inspection resolution reason');
     return value;
 }
 
diff --git a/web/js/workspace/prompt-blocks.mjs b/web/js/workspace/prompt-blocks.mjs
index f221b67..f3a3bcb 100644
--- a/web/js/workspace/prompt-blocks.mjs
+++ b/web/js/workspace/prompt-blocks.mjs
@@ -114,7 +114,7 @@ export function mountPromptBlocks(container, {prompts, sender, parentScope, role
                 heading.append(actions);
                 const status = el('p', '', 'prompt-copy-status muted'); status.setAttribute('role', 'status');
                 block.append(status);
-            } else block.append(el('p', typeof record?.message === 'string' ? record.message : statuses[record?.status] || statuses.unresolved, 'muted'));
+            } else block.append(el('p', typeof record?.message === 'string' && record.message ? record.message : statuses[record?.status] || statuses.unresolved, 'muted'));
             container.append(block);
         }
         if (includePairAction && Object.hasOwn(roles, 'positive') && Object.hasOwn(roles, 'negative')) {