diff --git a/mateclaw-server/src/main/java/vip/mate/agent/AgentGraphBuilder.java b/mateclaw-server/src/main/java/vip/mate/agent/AgentGraphBuilder.java
index bc725646..de3188b9 100644
--- a/mateclaw-server/src/main/java/vip/mate/agent/AgentGraphBuilder.java
+++ b/mateclaw-server/src/main/java/vip/mate/agent/AgentGraphBuilder.java
@@ -206,6 +206,17 @@ public class AgentGraphBuilder {
         Set<String> boundTools = agentBindingService.getEffectiveToolNames(entity.getId());
         toolSet = toolSet.withAllowedToolsOnly(boundTools); // null = 全局默认
 
+        // Issue #184 follow-up: an agent that opted out of skills must not be
+        // able to circle back and discover/load them via the meta tools. Strip
+        // the skill-discovery surface (listAvailableSkills / load_skill /
+        // readSkillFile / runSkillScript / listSkillFiles) here. This runs as a
+        // separate deny layer so the allowlist matrix in getEffectiveToolNames
+        // stays untouched — in particular, the (skillsDisabled, !toolsDisabled,
+        // no tool bindings) cell still returns null so non-skill global tools
+        // continue to flow through.
+        toolSet = toolSet.withDeniedToolsFiltered(
+                agentBindingService.getSkillDiscoveryDeniedTools(entity.getId()));
+
         // Escape hatch: drop the load_skill meta tool entirely when disabled, so
         // it isn't advertised regardless of binding (the catalog guidance falls
         // back to readSkillFile — see SkillRuntimeService).
diff --git a/mateclaw-server/src/main/java/vip/mate/agent/binding/service/AgentBindingService.java b/mateclaw-server/src/main/java/vip/mate/agent/binding/service/AgentBindingService.java
index 115ba089..d002e80b 100644
--- a/mateclaw-server/src/main/java/vip/mate/agent/binding/service/AgentBindingService.java
+++ b/mateclaw-server/src/main/java/vip/mate/agent/binding/service/AgentBindingService.java
@@ -583,6 +583,40 @@ public class AgentBindingService implements AgentBindingResolver {
         }
     }
 
+    /**
+     * Skill-discovery meta tools that let the LLM enumerate, load, read, or
+     * execute the workspace's skill catalog. Normally these live in
+     * {@link #SYSTEM_LEVEL_TOOLS} because every agent needs them — but when
+     * an agent has opted out of skills (issue #184), keeping them callable
+     * defeats the opt-out: the LLM can simply call {@code listAvailableSkills}
+     * to enumerate the catalog and {@code load_skill} to pull a SKILL.md
+     * into the conversation, even though the SKILL.md catalog itself was
+     * suppressed from the system prompt.
+     *
+     * <p>Resolved via {@link #getSkillDiscoveryDeniedTools} as a separate
+     * deny layer chained after the main allowlist, so the four-state matrix
+     * in {@link #getEffectiveToolNames} stays untouched.
+     */
+    private static final Set<String> SKILL_DISCOVERY_TOOLS = Set.of(
+            "listAvailableSkills",
+            "load_skill",
+            "readSkillFile",
+            "runSkillScript",
+            "listSkillFiles"
+    );
+
+    /**
+     * Tools the agent must NOT see when {@link AgentEntity#getSkillsDisabled()}
+     * is {@code true}. Empty otherwise. Chained on top of the allowlist by
+     * {@code AgentGraphBuilder} via {@code withDeniedToolsFiltered}.
+     */
+    public Set<String> getSkillDiscoveryDeniedTools(Long agentId) {
+        if (isSkillsDisabled(agentId)) {
+            return SKILL_DISCOVERY_TOOLS;
+        }
+        return Set.of();
+    }
+
     /**
      * Tools that exist outside the skill scope and must survive any
      * agent-level skill binding restriction.
diff --git a/mateclaw-server/src/test/java/vip/mate/agent/binding/AgentBindingServiceTest.java b/mateclaw-server/src/test/java/vip/mate/agent/binding/AgentBindingServiceTest.java
index f8c17552..0f3dc8ff 100644
--- a/mateclaw-server/src/test/java/vip/mate/agent/binding/AgentBindingServiceTest.java
+++ b/mateclaw-server/src/test/java/vip/mate/agent/binding/AgentBindingServiceTest.java
@@ -645,4 +645,62 @@ class AgentBindingServiceTest {
         assertFalse(readToolsDisabledFlag(),
                 "单条 bindTool 也算明确的承诺，应当自动清掉 flag");
     }
+
+    // ==================== Issue #184 follow-up: skill-discovery deny ====================
+
+    @Test
+    @DisplayName("issue #184 follow-up: skills_disabled=true → 屏蔽 listAvailableSkills / load_skill / readSkillFile / runSkillScript / listSkillFiles")
+    void skillsDisabledDeniesAllSkillDiscoveryTools() {
+        // Verified during smoke test: even with skillsDisabled=true the LLM
+        // could call listAvailableSkills and discover the full catalog. The
+        // deny layer below subtracts the 5 skill-discovery tools so the opt-out
+        // is honored end-to-end, not just in the SKILL.md catalog injection.
+        setSkillsDisabledFlag(true);
+
+        Set<String> denied = bindingService.getSkillDiscoveryDeniedTools(agentId);
+        assertEquals(5, denied.size(), "应当返回 5 个 skill-discovery 工具名");
+        assertTrue(denied.contains("listAvailableSkills"));
+        assertTrue(denied.contains("load_skill"));
+        assertTrue(denied.contains("readSkillFile"));
+        assertTrue(denied.contains("runSkillScript"));
+        assertTrue(denied.contains("listSkillFiles"));
+    }
+
+    @Test
+    @DisplayName("issue #184 follow-up: skills_disabled=false → 不屏蔽任何工具（保留向后兼容）")
+    void skillsEnabledReturnsEmptyDenySet() {
+        // Default state — no flag, no rows. The deny layer must be a no-op
+        // so a legacy agent keeps every skill-discovery tool it had before.
+        Set<String> denied = bindingService.getSkillDiscoveryDeniedTools(agentId);
+        assertNotNull(denied);
+        assertTrue(denied.isEmpty(),
+                "skillsDisabled=false 时 deny 集必须为空，否则会误伤未禁用技能的 agent");
+    }
+
+    @Test
+    @DisplayName("issue #184 follow-up: setSkillBindings 非空时 auto-clear flag → deny 也回到空集")
+    void skillDiscoveryDenyTracksAutoClearOfFlag() {
+        // Auto-clear contract from the main PR: writing a non-empty binding
+        // clears skills_disabled. The deny set must follow — it reads the
+        // same flag at call time, so after auto-clear it should be empty.
+        long skillId = 7_777_901L;
+        seedSkill(skillId);
+        setSkillsDisabledFlag(true);
+        assertEquals(5, bindingService.getSkillDiscoveryDeniedTools(agentId).size(),
+                "前置：flag 开启时 deny 应为 5 个");
+
+        bindingService.setSkillBindings(agentId, List.of(skillId));
+
+        assertTrue(bindingService.getSkillDiscoveryDeniedTools(agentId).isEmpty(),
+                "auto-clear 后 deny 必须回到空集，否则用户重新绑定技能后仍然看不到 listAvailableSkills");
+    }
+
+    @Test
+    @DisplayName("issue #184 follow-up: missing agent → deny 空集（防御性）")
+    void skillDiscoveryDenyHandlesMissingAgent() {
+        Set<String> denied = bindingService.getSkillDiscoveryDeniedTools(999_999_999L);
+        assertNotNull(denied);
+        assertTrue(denied.isEmpty(),
+                "agent 不存在时 deny 集应为空 —— 严格但不抛错，与 isSkillsDisabled 的契约一致");
+    }
 }