From 34d00fd323c793278c6ed126308c58e2fc7efd54 Mon Sep 17 00:00:00 2001
From: morluto <76467478+morluto@users.noreply.github.com>
Date: Tue, 29 Sep 2026 12:41:28 +0800
Subject: [PATCH] refactor(mcp): simplify workflow and observation scope

---
 docs/electron-observation.md                  | 16 +++-
 docs/javascript-runtime-observation.md        |  6 ++
 skills/reverse-engineer-anything/SKILL.md     |  8 +-
 .../references/native-and-artifacts.md        | 21 +++--
 src/contracts/electronToolContracts.ts        | 13 +--
 ...vascriptRuntimeObservationToolContracts.ts |  6 +-
 src/contracts/promptContracts.test.ts         | 11 ++-
 src/contracts/promptContracts.ts              | 17 +---
 src/domain/electronObservation.ts             |  8 +-
 src/domain/javascriptRuntimeObservation.ts    | 17 ++--
 src/server/createServer.ts                    | 12 ++-
 src/server/registerElectronTools.ts           | 85 ++++++++++++++++---
 ...gisterJavaScriptRuntimeObservationTools.ts | 77 +++++++++++++++--
 tests/boundary/mcp/electronMcp.test.ts        | 13 ++-
 .../javascriptRuntimeObservationMcp.test.ts   |  5 +-
 tests/boundary/mcp/serverWorkflowMcp.test.ts  |  5 +-
 16 files changed, 239 insertions(+), 81 deletions(-)

diff --git a/docs/electron-observation.md b/docs/electron-observation.md
index 0009eea0b7..902e8f98ee 100644
--- a/docs/electron-observation.md
+++ b/docs/electron-observation.md
@@ -28,6 +28,12 @@ Endpoints accept only explicit-port loopback HTTP URLs. Roots are canonicalized
 inline array. CDP response-size limits remain internal and oversized discovery
 responses fail explicitly.
 
+For the MCP follow-up call, pass the discovery Evidence ID and selected target
+ID to `inspect_electron_page`. REA reuses the endpoint and authorized roots from
+that retained discovery Evidence, checks the selected ID against its target
+list, and reapplies permission policy. The agent does not repeat the roots. The
+CLI continues to accept its endpoint and roots directly.
+
 ```bash
 rea list-electron-targets http://127.0.0.1:9223 --json
 rea inspect-electron-page http://127.0.0.1:9223 TARGET_ID \
@@ -93,16 +99,18 @@ The CLI and MCP surfaces accept the same schema. For a JSON request file:
 rea capture-electron-scenario scenario.json --json
 ```
 
-The result records bounded action status and targets, correlated app/window/WebContents,
+The result records action status and targets, correlated app/window/WebContents,
 preload, session, navigation, shell, permission, popup, download, protocol,
 native-addon, process, and IPC timeline events. IPC channels are capped at
 1,024 characters and argument-shape metadata at 32 entries; values are never
-retained. Renderer crash/restart and deep-link actions are synthetic, bounded
+retained. Renderer crash/restart and deep-link actions are synthetic
 scenario controls; their attempted and observed outcomes remain in the
 timeline. The active hook blocks and records external shell/navigation,
 permission, download, popup, updater, and OS-integration effects. The timeline
-is explicitly partial when attachment starts after application activity or when
-the event budget is exhausted.
+is explicitly partial when attachment starts after application activity. The
+owned experiment has an internal 60-second deadline and a 5-second per-action
+timeout so hung work reaches process cleanup; these are lifecycle timeouts, not
+caller-selected capture limits.
 
 Active Electron Evidence can also be supplied to
 [`reconcile_javascript_runtime`](javascript-runtime-reconciliation.md). That
diff --git a/docs/javascript-runtime-observation.md b/docs/javascript-runtime-observation.md
index 4288ac842b..3512b4dcd2 100644
--- a/docs/javascript-runtime-observation.md
+++ b/docs/javascript-runtime-observation.md
@@ -35,6 +35,12 @@ Excluded locations are counted but never retained.
 array. The Inspector discovery response-size ceiling remains internal; an
 oversized response fails explicitly.
 
+For the MCP follow-up call, pass the discovery Evidence ID and selected target
+ID to `observe_javascript_runtime`. REA reuses the endpoint and file/origin
+scope from that retained discovery Evidence, checks the selected ID against its
+target list, and reapplies permission policy. The agent does not repeat roots or
+origins. The CLI continues to accept its endpoint and scope arguments directly.
+
 ## Passive protocol boundary
 
 The provider sends exactly two protocol commands:
diff --git a/skills/reverse-engineer-anything/SKILL.md b/skills/reverse-engineer-anything/SKILL.md
index 55b2e63961..15b0eda1ab 100644
--- a/skills/reverse-engineer-anything/SKILL.md
+++ b/skills/reverse-engineer-anything/SKILL.md
@@ -4,7 +4,7 @@ description: Reverse engineer native, managed, Electron/JavaScript, packaged, an
 metadata:
   version: "23"
   tool_count: 111
-  catalog_digest: "8e57d3a38cf7e8873ccf94037adb5f59cf1eceba8dc18bd20c95baff19689ebf"
+  catalog_digest: "fcd5f0637e95ab25da4598c2e6c596cebedbc039efe43521fff6ef982e4a6b2a"
 ---
 
 # REA
@@ -23,12 +23,14 @@ Choose the first tool from the target the user supplied. Do not call
 - ASAR or extracted JavaScript/Electron tree:
   `analyze_javascript_application`.
 - Archive, application package, ZIP/APK/IPA/MSIX/AppX, or DMG:
-  `open_binary` with the supplied local path, then `inspect_artifact`.
+  `open_binary` with the supplied local path; use `inspect_artifact` when its
+  graph and findings help answer the question.
 - Managed PE/CLI assembly: `inspect_managed_artifact`.
 - User-owned browser page already open: `list_browser_targets`.
 - User-owned Electron runtime already open: `list_electron_targets`.
 - Native executable, library, or analysis database: `open_binary`, then
-  `binary_overview`.
+  use focused analysis tools directly; call `binary_overview` when metadata or
+  inventory context is useful.
 
 If the app is missing, ask which app to inspect. Resolve a human-readable app
 name to one clear installed artifact when possible; ask only when matches are
diff --git a/skills/reverse-engineer-anything/references/native-and-artifacts.md b/skills/reverse-engineer-anything/references/native-and-artifacts.md
index bcac20b712..9fe0afdcb9 100644
--- a/skills/reverse-engineer-anything/references/native-and-artifacts.md
+++ b/skills/reverse-engineer-anything/references/native-and-artifacts.md
@@ -2,8 +2,9 @@
 
 ## Native targets
 
-After `open_binary`, use `binary_overview` once and narrow around the requested
-feature. Prefer literal search, names, decompilation, callers, callees, and
+After `open_binary`, use focused search, procedure, or function tools directly.
+Use `binary_overview` when target metadata or inventory context helps answer the
+question. Prefer literal search, names, decompilation, callers, callees, and
 cross-references. Addresses and recovered pseudocode are analysis observations,
 not original source. Provider unavailability and unsupported metadata remain
 unknown rather than false.
@@ -23,14 +24,16 @@ parser and must not become an implicit setup dependency.
 
 ## Packages and extraction
 
-Call `inspect_artifact` before extraction for application bundles, archives,
-ZIP/APK/IPA/MSIX/AppX, ASAR, or DMG inputs. It returns the complete artifact
-graph inline. Cite graph manifest IDs.
+Use `inspect_artifact` for application bundles, archives, ZIP/APK/IPA/MSIX/AppX,
+ASAR, or DMG inputs when the artifact graph and findings help answer the
+question. It returns the complete artifact graph inline. Cite graph manifest
+IDs when using them.
 
-`extract_artifact` takes no arguments and materializes all regular files into a
-fresh temporary directory chosen by REA. Symlinks and encrypted entries are
-inventory facts, not extractable files. The operation still requires the
-filesystem-write permission grant; callers do not choose the destination.
+Use `extract_artifact` when materialized files are needed. It takes no arguments
+and materializes all regular files into a fresh temporary directory chosen by
+REA. Symlinks and encrypted entries are inventory facts, not extractable files.
+The operation still requires the filesystem-write permission grant; callers do
+not choose the destination.
 
 Native DMG traversal is macOS-only, read-only, and requires both operator policy
 and `native_mount_approved: true`; without both, retain the root-hash-only result.
diff --git a/src/contracts/electronToolContracts.ts b/src/contracts/electronToolContracts.ts
index 7f89e8d34e..4a8fa120ec 100644
--- a/src/contracts/electronToolContracts.ts
+++ b/src/contracts/electronToolContracts.ts
@@ -38,14 +38,6 @@ const activeExample = {
   application_root: "/Applications/Example.app/Contents/Resources",
   args: [],
   actions: [{ step_id: "exercise-ipc", kind: "click", selector: "#run" }],
-  limits: {
-    max_duration_ms: 60_000,
-    action_timeout_ms: 5_000,
-    max_actions: 20,
-    max_ipc_events: 2_000,
-    max_processes: 32,
-    max_windows: 32,
-  },
 };
 
 /** Root-confined Electron file-page discovery and inspection contracts. */
@@ -72,7 +64,7 @@ export const ELECTRON_TOOL_CONTRACTS = [
     name: "inspect_electron_page",
     ...toolContractMetadata("inspect_electron_page"),
     description:
-      "Passively inspect one approved Electron file page through CDP. Returns root-confined frames, DOM structure, resource paths, and scripts without evaluating renderer JavaScript or invoking Electron APIs; script contents require separate approval.",
+      "Passively inspect one Electron file page using its list_electron_targets Evidence and returned target ID. Pass the discovery Evidence ID to reuse its endpoint and authorized file roots; this avoids repeating scope and cannot widen permission. Returns frames, DOM structure, resource paths, and scripts without evaluating renderer JavaScript or invoking Electron APIs.",
     kind: "electron-provider",
     inputSchema: inspectElectronPageToolInputSchema,
     outputSchema: inspectionOutputSchema,
@@ -80,8 +72,7 @@ export const ELECTRON_TOOL_CONTRACTS = [
       {
         title: "Inspect an approved Electron page",
         input: {
-          cdp_endpoint: endpoint,
-          allowed_file_roots: [root],
+          discovery_evidence_id: "EVIDENCE_ID_FROM_LIST_ELECTRON_TARGETS",
           target_id: "TARGET_ID_FROM_LIST_ELECTRON_TARGETS",
         },
       },
diff --git a/src/contracts/javascriptRuntimeObservationToolContracts.ts b/src/contracts/javascriptRuntimeObservationToolContracts.ts
index e7fbb5bf2d..da8387eae3 100644
--- a/src/contracts/javascriptRuntimeObservationToolContracts.ts
+++ b/src/contracts/javascriptRuntimeObservationToolContracts.ts
@@ -36,7 +36,7 @@ export const JAVASCRIPT_RUNTIME_OBSERVATION_TOOL_CONTRACTS = [
     name: "observe_javascript_runtime",
     ...toolContractMetadata("observe_javascript_runtime"),
     description:
-      "Attach passively to one exact approved Node/Electron V8 Inspector target and capture bounded Debugger.scriptParsed plus Runtime execution-context events. REA sends only Runtime.enable and Debugger.enable: it never evaluates, pauses, resumes, reads source, or instruments the target. require/import edges, EventEmitter activity, and Electron IPC remain explicit unknowns; reconcile the returned Evidence with static Application Graph Evidence using reconcile_javascript_runtime.",
+      "Attach passively to one Node/Electron V8 Inspector target using its list_javascript_runtime_targets Evidence and returned target ID. Pass the discovery Evidence ID to reuse its endpoint and authorized file/origin scope; this avoids repeating scope and cannot widen permission. Captures Debugger.scriptParsed and Runtime execution-context events. REA never evaluates, pauses, resumes, reads source, or instruments the target; require/import edges, EventEmitter activity, and Electron IPC remain explicit unknowns. Reconcile the result with static Application Graph Evidence using reconcile_javascript_runtime.",
     kind: "runtime-provider",
     inputSchema: observeJavaScriptRuntimeToolInputSchema,
     outputSchema: evidenceResultOf(javascriptRuntimeObservationSchema),
@@ -44,8 +44,8 @@ export const JAVASCRIPT_RUNTIME_OBSERVATION_TOOL_CONTRACTS = [
       {
         title: "Observe one approved Node runtime",
         input: {
-          inspector_endpoint: endpoint,
-          allowed_file_roots: [root],
+          discovery_evidence_id:
+            "EVIDENCE_ID_FROM_LIST_JAVASCRIPT_RUNTIME_TARGETS",
           target_id: "TARGET_ID_FROM_LIST_JAVASCRIPT_RUNTIME_TARGETS",
         },
       },
diff --git a/src/contracts/promptContracts.test.ts b/src/contracts/promptContracts.test.ts
index 181c120929..ac3131b57d 100644
--- a/src/contracts/promptContracts.test.ts
+++ b/src/contracts/promptContracts.test.ts
@@ -84,15 +84,14 @@ describe("guided prompt contracts", () => {
     ).toBeLessThan(
       rendered.get("audit_residual_unknowns")?.indexOf("`update_unknown`") ?? 0,
     );
+    expect(rendered.get("prepare_bounded_process_capture")).not.toContain(
+      "`binary_session`",
+    );
     expect(
       rendered
         .get("prepare_bounded_process_capture")
-        ?.indexOf("`binary_session`"),
-    ).toBeLessThan(
-      rendered
-        .get("prepare_bounded_process_capture")
-        ?.indexOf("`capture_process_scenario`") ?? 0,
-    );
+        ?.indexOf("`capture_process_scenario`"),
+    ).toBeGreaterThanOrEqual(0);
     expect(
       rendered
         .get("compare_application_versions")
diff --git a/src/contracts/promptContracts.ts b/src/contracts/promptContracts.ts
index 048c5beeec..a0d2bbabfa 100644
--- a/src/contracts/promptContracts.ts
+++ b/src/contracts/promptContracts.ts
@@ -116,18 +116,14 @@ export const PROMPT_CONTRACTS = [
     },
     steps: [
       {
-        tools: [
-          "binary_session",
-          "open_binary",
-          "analyze_javascript_application",
-        ],
+        tools: ["open_binary", "analyze_javascript_application"],
         instruction:
           "For an approved JavaScript/Electron artifact, reconstruct its application graph without execution when useful; open target_path only when no matching target is active.",
       },
       {
-        tools: ["list_documents", "set_current_document", "binary_overview"],
+        tools: ["list_documents", "set_current_document"],
         instruction:
-          "Select an explicitly discovered document, then establish bounded binary, language, segment, and procedure context.",
+          "Select a document only when needed for the requested analysis.",
       },
       {
         tools: ["search_strings", "search_procedures", "procedure_address"],
@@ -294,7 +290,7 @@ export const PROMPT_CONTRACTS = [
     },
     steps: [
       {
-        tools: ["binary_session", "open_binary", "binary_overview"],
+        tools: ["open_binary"],
         instruction:
           "Confirm the target only when needed; open or analyze it if it is not already active and the request requires it.",
       },
@@ -395,11 +391,6 @@ export const PROMPT_CONTRACTS = [
       ),
     },
     steps: [
-      {
-        tools: ["binary_session"],
-        instruction:
-          "Inspect capture capability, declared effects, and limits before proposing execution. An unavailable capability is not permission to widen policy.",
-      },
       {
         tools: ["get_evidence_bundle"],
         instruction:
diff --git a/src/domain/electronObservation.ts b/src/domain/electronObservation.ts
index 7e867a1562..6b810de39f 100644
--- a/src/domain/electronObservation.ts
+++ b/src/domain/electronObservation.ts
@@ -66,11 +66,15 @@ export const inspectElectronPageInputSchema = z
   }));
 export const inspectElectronPageToolInputSchema = z.union([
   z.strictObject({
-    ...inspectElectronPageFacts,
+    discovery_evidence_id: z.string().min(1),
+    target_id: z.string().trim().min(1).max(256),
+    observation_ms: z.number().int().min(0).max(10_000).default(100),
     include_script_sources: z.literal(false).default(false),
   }),
   z.strictObject({
-    ...inspectElectronPageFacts,
+    discovery_evidence_id: z.string().min(1),
+    target_id: z.string().trim().min(1).max(256),
+    observation_ms: z.number().int().min(0).max(10_000).default(100),
     include_script_sources: z.literal(true),
   }),
 ]);
diff --git a/src/domain/javascriptRuntimeObservation.ts b/src/domain/javascriptRuntimeObservation.ts
index 0e514ad449..f37ac209a9 100644
--- a/src/domain/javascriptRuntimeObservation.ts
+++ b/src/domain/javascriptRuntimeObservation.ts
@@ -82,18 +82,23 @@ export const DEFAULT_JAVASCRIPT_RUNTIME_OBSERVATION_LIMITS = {
 };
 
 /** Input for one bounded, attach-only V8 Inspector observation. */
-export const observeJavaScriptRuntimeToolInputSchema = z
+export const observeJavaScriptRuntimeToolInputSchema = z.strictObject({
+  discovery_evidence_id: z.string().min(1),
+  target_id: z.string().trim().min(1).max(256),
+  runtime_kind: javascriptRuntimeKindSchema.optional(),
+  observation_ms: z.number().int().min(0).max(10_000).default(100),
+});
+
+/** Input after discovery Evidence supplies the authorized target scope. */
+export const observeJavaScriptRuntimeInputSchema = z
   .strictObject({
     ...runtimeScope,
     target_id: z.string().trim().min(1).max(256),
     runtime_kind: javascriptRuntimeKindSchema.optional(),
     observation_ms: z.number().int().min(0).max(10_000).default(100),
   })
-  .superRefine(requireRuntimeScope);
-
-/** Internal parsed input with REA's fixed collection bounds applied. */
-export const observeJavaScriptRuntimeInputSchema =
-  observeJavaScriptRuntimeToolInputSchema.transform((input) => ({
+  .superRefine(requireRuntimeScope)
+  .transform((input) => ({
     ...input,
     limits: DEFAULT_JAVASCRIPT_RUNTIME_OBSERVATION_LIMITS,
   }));
diff --git a/src/server/createServer.ts b/src/server/createServer.ts
index e726b852b3..ddb4a0ef92 100644
--- a/src/server/createServer.ts
+++ b/src/server/createServer.ts
@@ -55,10 +55,10 @@ import {
 } from "./sessionToolPolicies.js";
 
 const TARGET_FREE_INSTRUCTIONS =
-  "Start with binary_session to see the active target and available providers. For an archive or application package, call open_binary(path), then inspect_artifact for the complete graph and findings; use extract_artifact to materialize its regular contents when needed. For standalone JavaScript or ASAR analysis, use analyze_javascript_application. For managed PE/CLI files, use inspect_managed_artifact. For native binaries, open the target and start with binary_overview. For a live page or runtime, discover targets with list_browser_targets, list_electron_targets, or list_javascript_runtime_targets before inspection; scenario capture tools are for controlled interactions.";
+  "Choose tools from the target and question. For an archive or application package, open_binary(path) binds the active target; use inspect_artifact when its graph and findings help, and extract_artifact when materialized files are needed. For standalone JavaScript or ASAR analysis, use analyze_javascript_application. For managed PE/CLI files, use inspect_managed_artifact. For native binaries, open the target and use focused search, procedure, or function tools directly; binary_overview is available when metadata or inventory context is useful. For a live page or runtime, discover targets with list_browser_targets, list_electron_targets, or list_javascript_runtime_targets when needed; scenario capture tools are for controlled interactions.";
 
 const ACTIVE_TARGET_INSTRUCTIONS =
-  "REA analyzes the active reverse-engineering target. For native binaries, call binary_overview for target metadata and inventory counts, then search_procedures or list_procedures to choose a function and analyze_function with its returned name or address for a complete dossier. Use inspect_native_api when you need to reconstruct a native API boundary, and focused procedure tools when you need only one facet such as assembly, callers, or references.";
+  "REA analyzes the active reverse-engineering target. For native binaries, use search_procedures or list_procedures to choose a function and analyze_function with its returned name or address for a complete dossier. Call binary_overview when target metadata or inventory counts are useful. Use inspect_native_api when you need to reconstruct a native API boundary, and focused procedure tools when you need only one facet such as assembly, callers, or references.";
 
 export interface CreateServerOptions {
   readonly logger?: Logger;
@@ -336,10 +336,18 @@ const registerObservationTools = ({
     ...common,
     electron: options.electronObservation,
     electronActive: options.electronActiveObservation,
+    evidenceLookup:
+      session === undefined
+        ? undefined
+        : (evidenceId) => session.evidenceById(evidenceId),
   });
   registerJavaScriptRuntimeObservationTools(server, {
     ...common,
     runtime: options.javascriptRuntimeObservation,
+    evidenceLookup:
+      session === undefined
+        ? undefined
+        : (evidenceId) => session.evidenceById(evidenceId),
   });
   registerApplicationTools(server, {
     logger,
diff --git a/src/server/registerElectronTools.ts b/src/server/registerElectronTools.ts
index 259e8c74a5..58361d5d6f 100644
--- a/src/server/registerElectronTools.ts
+++ b/src/server/registerElectronTools.ts
@@ -15,9 +15,16 @@ import type { ProgressReporter } from "../application/ProgressReporter.js";
 import { ELECTRON_TOOL_CONTRACTS } from "../contracts/electronToolContracts.js";
 import type { ToolContract } from "../contracts/toolContracts.js";
 import type { AnalysisError } from "../domain/errors.js";
+import { AnalysisInputError } from "../domain/errors.js";
+import { resolveEvidenceReferences } from "../application/EvidenceReferenceResolver.js";
 import type { Evidence } from "../domain/evidence.js";
-import type { Result } from "../domain/result.js";
-import { inspectElectronPageInputSchema } from "../domain/electronObservation.js";
+import { err, type Result } from "../domain/result.js";
+import {
+  electronTargetListSchema,
+  inspectElectronPageInputSchema,
+  inspectElectronPageToolInputSchema,
+  listElectronTargetsInputSchema,
+} from "../domain/electronObservation.js";
 import type { Logger } from "../logger.js";
 import { mcpProgressReporter } from "./mcpProgress.js";
 import { logToolExecution } from "./toolLogging.js";
@@ -30,6 +37,9 @@ interface ElectronToolRegistration {
   readonly electronActive: ElectronActiveObservationPort | undefined;
   readonly permissionAuthority: PermissionAuthority | undefined;
   readonly recordEvidence: BinarySessionPort["recordEvidence"] | undefined;
+  readonly evidenceLookup:
+    | ((evidenceId: string) => Evidence | undefined)
+    | undefined;
 }
 
 interface ElectronToolContext {
@@ -78,16 +88,61 @@ export const registerElectronTools = (
         options,
         inspectContract,
         { input, context },
-        (parsed, { signal, progress }) =>
-          inspectElectronPage(
+        async (parsed, { signal, progress }) => {
+          const request = inspectElectronPageToolInputSchema.parse(parsed);
+          const source = resolveEvidenceReferences(
+            options.evidenceLookup,
+            [request.discovery_evidence_id],
+            [
+              {
+                operation: "list_electron_targets",
+                predicate: "rea.electron-target-list/v1",
+              },
+            ],
+          );
+          if (!source.ok) return err(source.error);
+          const discovery = source.value[0];
+          if (discovery === undefined) return err(invalidDiscoveryEvidence());
+          const scope = listElectronTargetsInputSchema.safeParse(
+            discovery.parameters,
+          );
+          const listed = electronTargetListSchema.safeParse(
+            discovery.normalized_result,
+          );
+          if (!scope.success || !listed.success)
+            return err(invalidDiscoveryEvidence());
+          if (
+            !listed.data.targets.some(
+              ({ target_id }) => target_id === request.target_id,
+            )
+          )
+            return {
+              ok: false,
+              error: new AnalysisInputError(
+                "inspect_electron_page",
+                undefined,
+                [
+                  {
+                    path: ["target_id"],
+                    reason: "invalid_value",
+                    message:
+                      "Choose a target returned by the referenced discovery Evidence",
+                  },
+                ],
+              ),
+            };
+          return inspectElectronPage(
             options.electron,
             options.permissionAuthority,
-            inspectElectronPageInputSchema.parse(parsed),
-            {
-              signal,
-              progress,
-            },
-          ),
+            inspectElectronPageInputSchema.parse({
+              ...scope.data,
+              target_id: request.target_id,
+              observation_ms: request.observation_ms,
+              include_script_sources: request.include_script_sources,
+            }),
+            { signal, progress },
+          );
+        },
       ),
   );
   server.registerTool(
@@ -133,6 +188,16 @@ export const registerElectronTools = (
   );
 };
 
+const invalidDiscoveryEvidence = () =>
+  new AnalysisInputError("inspect_electron_page", undefined, [
+    {
+      path: ["discovery_evidence_id"],
+      reason: "invalid_value",
+      message:
+        "The referenced Evidence is not a usable Electron target discovery",
+    },
+  ]);
+
 const runElectronTool = async <Input>(
   options: ElectronToolRegistration,
   contract: ToolContract,
diff --git a/src/server/registerJavaScriptRuntimeObservationTools.ts b/src/server/registerJavaScriptRuntimeObservationTools.ts
index 81ddf7df33..7b036552ce 100644
--- a/src/server/registerJavaScriptRuntimeObservationTools.ts
+++ b/src/server/registerJavaScriptRuntimeObservationTools.ts
@@ -7,12 +7,19 @@ import {
   observeJavaScriptRuntime,
 } from "../application/JavaScriptRuntimeObservationService.js";
 import type { PermissionAuthority } from "../application/PermissionAuthority.js";
+import { resolveEvidenceReferences } from "../application/EvidenceReferenceResolver.js";
 import { JAVASCRIPT_RUNTIME_OBSERVATION_TOOL_CONTRACTS } from "../contracts/javascriptRuntimeObservationToolContracts.js";
 import type { ToolContract } from "../contracts/toolContracts.js";
 import type { AnalysisError } from "../domain/errors.js";
+import { AnalysisInputError } from "../domain/errors.js";
 import type { Evidence } from "../domain/evidence.js";
-import type { Result } from "../domain/result.js";
-import { observeJavaScriptRuntimeInputSchema } from "../domain/javascriptRuntimeObservation.js";
+import { err, type Result } from "../domain/result.js";
+import {
+  javascriptRuntimeTargetListSchema,
+  listJavaScriptRuntimeTargetsInputSchema,
+  observeJavaScriptRuntimeInputSchema,
+  observeJavaScriptRuntimeToolInputSchema,
+} from "../domain/javascriptRuntimeObservation.js";
 import type { Logger } from "../logger.js";
 import { logToolExecution } from "./toolLogging.js";
 import { toolRegistrationOptions } from "./toolRegistrationOptions.js";
@@ -23,6 +30,9 @@ interface RuntimeToolRegistration {
   readonly runtime: JavaScriptRuntimeObservationPort | undefined;
   readonly permissionAuthority: PermissionAuthority | undefined;
   readonly recordEvidence: BinarySessionPort["recordEvidence"] | undefined;
+  readonly evidenceLookup:
+    | ((evidenceId: string) => Evidence | undefined)
+    | undefined;
 }
 
 /** Register passive Inspector tools even when policy keeps them unavailable. */
@@ -57,17 +67,72 @@ export const registerJavaScriptRuntimeObservationTools = (
         options,
         observeContract,
         { input, context },
-        (parsed, signal) =>
-          observeJavaScriptRuntime(
+        async (parsed, signal) => {
+          const request = observeJavaScriptRuntimeToolInputSchema.parse(parsed);
+          const source = resolveEvidenceReferences(
+            options.evidenceLookup,
+            [request.discovery_evidence_id],
+            [
+              {
+                operation: "list_javascript_runtime_targets",
+                predicate: "rea.javascript-runtime-target-list/v1",
+              },
+            ],
+          );
+          if (!source.ok) return err(source.error);
+          const discovery = source.value[0];
+          if (discovery === undefined) return err(invalidDiscoveryEvidence());
+          const scope = listJavaScriptRuntimeTargetsInputSchema.safeParse(
+            discovery.parameters,
+          );
+          const listed = javascriptRuntimeTargetListSchema.safeParse(
+            discovery.normalized_result,
+          );
+          if (!scope.success || !listed.success)
+            return err(invalidDiscoveryEvidence());
+          if (
+            !listed.data.targets.some(
+              ({ target_id }) => target_id === request.target_id,
+            )
+          )
+            return err(
+              new AnalysisInputError("observe_javascript_runtime", undefined, [
+                {
+                  path: ["target_id"],
+                  reason: "invalid_value",
+                  message:
+                    "Choose a target returned by the referenced discovery Evidence",
+                },
+              ]),
+            );
+          return observeJavaScriptRuntime(
             options.runtime,
             options.permissionAuthority,
-            observeJavaScriptRuntimeInputSchema.parse(parsed),
+            observeJavaScriptRuntimeInputSchema.parse({
+              ...scope.data,
+              target_id: request.target_id,
+              ...(request.runtime_kind === undefined
+                ? {}
+                : { runtime_kind: request.runtime_kind }),
+              observation_ms: request.observation_ms,
+            }),
             { signal },
-          ),
+          );
+        },
       ),
   );
 };
 
+const invalidDiscoveryEvidence = () =>
+  new AnalysisInputError("observe_javascript_runtime", undefined, [
+    {
+      path: ["discovery_evidence_id"],
+      reason: "invalid_value",
+      message:
+        "The referenced Evidence is not a usable Inspector target discovery",
+    },
+  ]);
+
 const runRuntimeTool = async <Input>(
   options: RuntimeToolRegistration,
   contract: ToolContract,
diff --git a/tests/boundary/mcp/electronMcp.test.ts b/tests/boundary/mcp/electronMcp.test.ts
index afddfc78f7..8b9cd43526 100644
--- a/tests/boundary/mcp/electronMcp.test.ts
+++ b/tests/boundary/mcp/electronMcp.test.ts
@@ -96,11 +96,11 @@ it("exposes root-confined Electron discovery and inspection as Evidence", async
       targets: [{ target_id: "electron-page" }],
     },
   });
+  const discoveryEvidenceId = evidenceIdFrom(listed.structuredContent);
   const inspected = await client.callTool({
     name: "inspect_electron_page",
     arguments: {
-      cdp_endpoint: browser.endpoint,
-      allowed_file_roots: [root],
+      discovery_evidence_id: discoveryEvidenceId,
       target_id: "electron-page",
       observation_ms: 0,
       include_script_sources: true,
@@ -369,3 +369,12 @@ const evidenceFor = (session: BinarySession, value: unknown) => {
   if (evidence === undefined) throw new TypeError("Missing session Evidence");
   return evidence;
 };
+
+const evidenceIdFrom = (value: unknown): string => {
+  if (typeof value !== "object" || value === null)
+    throw new TypeError("Missing structured result");
+  const evidenceId = Reflect.get(value, "evidence_id");
+  if (typeof evidenceId !== "string")
+    throw new TypeError("Missing Evidence ID");
+  return evidenceId;
+};
diff --git a/tests/boundary/mcp/javascriptRuntimeObservationMcp.test.ts b/tests/boundary/mcp/javascriptRuntimeObservationMcp.test.ts
index a7e3742a0d..4a25bb4f04 100644
--- a/tests/boundary/mcp/javascriptRuntimeObservationMcp.test.ts
+++ b/tests/boundary/mcp/javascriptRuntimeObservationMcp.test.ts
@@ -81,13 +81,12 @@ describe("JavaScript runtime observation MCP tools", () => {
         targets: [{ target_id: inspector.targetId }],
       },
     });
+    const discoveryEvidenceId = evidenceIdFrom(listed.structuredContent);
 
     const observedRuntime = await client.callTool({
       name: "observe_javascript_runtime",
       arguments: {
-        inspector_endpoint: inspector.endpoint,
-        allowed_file_roots: [root],
-        allowed_origins: [],
+        discovery_evidence_id: discoveryEvidenceId,
         target_id: inspector.targetId,
         runtime_kind: "node",
         observation_ms: 10,
diff --git a/tests/boundary/mcp/serverWorkflowMcp.test.ts b/tests/boundary/mcp/serverWorkflowMcp.test.ts
index 48b2afb158..6394d17283 100644
--- a/tests/boundary/mcp/serverWorkflowMcp.test.ts
+++ b/tests/boundary/mcp/serverWorkflowMcp.test.ts
@@ -49,7 +49,10 @@ it("advertises the native function discovery-to-dossier workflow", async () => {
   });
 
   expect(client.getInstructions()).toContain(
-    "call binary_overview for target metadata and inventory counts, then search_procedures or list_procedures to choose a function",
+    "use search_procedures or list_procedures to choose a function",
+  );
+  expect(client.getInstructions()).toContain(
+    "Call binary_overview when target metadata or inventory counts are useful",
   );
 
   const tools = new Map(