fix(growth,skills,kernel): Phase 0 地基修复 — 经验积累覆盖 + Skill 工具调用

Bug 1: ExperienceStore store_experience() 相同 pain_pattern 因确定性 URI
直接覆盖，新 Experience reuse_count=0 重置已有积累。修复为先检查 URI
是否已存在，若存在则合并（保留原 id/created_at，reuse_count+1）。

Bug 2: PromptOnlySkill::execute() 只做纯文本 complete()，75 个 Skill
的 tools 字段是装饰性的。修复为扩展 LlmCompleter 支持 complete_with_tools，
SkillContext 新增 tool_definitions，KernelSkillExecutor 从 ToolRegistry
解析 manifest 声明的工具定义传入 LLM function calling。

crates/zclaw-skills/src/runner.rs

@@ -7,7 +7,7 @@ use std::time::Instant;
 use tracing::warn;
 use zclaw_types::Result;
 
-use super::{Skill, SkillContext, SkillManifest, SkillResult};
+use super::{Skill, SkillCompletion, SkillContext, SkillManifest, SkillResult};
 
 /// Returns the platform-appropriate Python binary name.
 /// On Windows, the standard installer provides `python.exe`, not `python3.exe`.
@@ -39,6 +39,17 @@ impl PromptOnlySkill {
 
         prompt
     }
+
+    fn completion_to_result(&self, completion: SkillCompletion) -> SkillResult {
+        if completion.tool_calls.is_empty() {
+            return SkillResult::success(Value::String(completion.text));
+        }
+        // Include both text and tool calls so the caller can relay them.
+        SkillResult::success(serde_json::json!({
+            "text": completion.text,
+            "tool_calls": completion.tool_calls,
+        }))
+    }
 }
 
 #[async_trait]
@@ -50,13 +61,25 @@ impl Skill for PromptOnlySkill {
     async fn execute(&self, context: &SkillContext, input: Value) -> Result<SkillResult> {
         let prompt = self.format_prompt(&input);
 
-        // If an LLM completer is available, generate an AI response
         if let Some(completer) = &context.llm {
+            // If tool definitions are available and the manifest declares tools,
+            // use tool-augmented completion so the LLM can invoke tools.
+            if !context.tool_definitions.is_empty() && !self.manifest.tools.is_empty() {
+                match completer.complete_with_tools(&prompt, None, context.tool_definitions.clone()).await {
+                    Ok(completion) => {
+                        return Ok(self.completion_to_result(completion));
+                    }
+                    Err(e) => {
+                        warn!("[PromptOnlySkill] Tool completion failed: {}, falling back", e);
+                    }
+                }
+            }
+
+            // Plain completion (no tools or fallback)
             match completer.complete(&prompt).await {
                 Ok(response) => return Ok(SkillResult::success(Value::String(response))),
                 Err(e) => {
                     warn!("[PromptOnlySkill] LLM completion failed: {}, falling back to raw prompt", e);
-                    // Fall through to return raw prompt
                 }
             }
         }