diff --git a/tools/http-framework-test.yaml b/tools/http-framework-test.yaml
index 20d777d3..eea4ff08 100644
--- a/tools/http-framework-test.yaml
+++ b/tools/http-framework-test.yaml
@@ -440,6 +440,230 @@ args:
             print("Body: <empty>")
 
 
+    def compile_response_filter(pattern: str, ignore_case: bool):
+        flags = 0
+        if ignore_case:
+            flags |= re.IGNORECASE
+        try:
+            return re.compile(pattern, flags)
+        except re.error as exc:
+            print(f"Invalid response_filter regex: {exc}", file=sys.stderr)
+            sys.exit(2)
+
+
+    def truncate_utf8(text: str, max_bytes: int) -> Tuple[str, bool]:
+        if max_bytes <= 0 or not text:
+            return text, False
+        encoded = text.encode("utf-8", errors="replace")
+        if len(encoded) <= max_bytes:
+            return text, False
+        truncated = encoded[:max_bytes].decode("utf-8", errors="ignore")
+        return truncated, True
+
+
+    def cap_line_entries(entries: List[Tuple[int, str]], max_lines: int) -> Tuple[List[Tuple[int, str]], bool]:
+        if max_lines <= 0 or len(entries) <= max_lines:
+            return entries, False
+        return entries[:max_lines], True
+
+
+    def expand_line_context(line_numbers: List[int], total_lines: int, context: int) -> List[int]:
+        if context <= 0:
+            return sorted(set(line_numbers))
+        included = set()
+        for num in line_numbers:
+            start = max(1, num - context)
+            end = min(total_lines, num + context)
+            for i in range(start, end + 1):
+                included.add(i)
+        return sorted(included)
+
+
+    def format_line_entries(lines: List[str], indices: List[int], ellipsis_gaps: bool = True) -> str:
+        if not indices:
+            return ""
+        chunks = []
+        prev = None
+        for num in indices:
+            if ellipsis_gaps and prev is not None and num > prev + 1:
+                chunks.append("  ...")
+            chunks.append(f"  L{num}: {lines[num - 1]}")
+            prev = num
+        return "\n".join(chunks)
+
+
+    def filter_body_by_lines(
+        lines: List[str],
+        compiled: "re.Pattern",
+        invert: bool,
+        context_lines: int,
+        max_lines: int,
+    ) -> Tuple[str, Dict[str, object]]:
+        matched_nums = []
+        for idx, line in enumerate(lines, start=1):
+            hit = compiled.search(line) is not None
+            if invert:
+                hit = not hit
+            if hit:
+                matched_nums.append(idx)
+        total = len(lines)
+        meta = {
+            "mode": "line",
+            "total_lines": total,
+            "matched_lines": len(matched_nums),
+            "invert": invert,
+            "truncated": False,
+            "byte_truncated": False,
+        }
+        if not matched_nums:
+            return "", meta
+        display_nums = expand_line_context(matched_nums, total, context_lines)
+        entries = [(n, lines[n - 1]) for n in display_nums]
+        entries, line_capped = cap_line_entries(entries, max_lines)
+        meta["truncated"] = line_capped
+        meta["display_lines"] = len(entries)
+        return format_line_entries(lines, [n for n, _ in entries], ellipsis_gaps=context_lines > 0), meta
+
+
+    def filter_body_multiline(
+        text: str,
+        compiled: "re.Pattern",
+        invert: bool,
+        max_lines: int,
+        dotall: bool,
+    ) -> Tuple[str, Dict[str, object]]:
+        flags = compiled.flags
+        if dotall:
+            pattern = re.compile(compiled.pattern, flags | re.DOTALL | re.MULTILINE)
+        else:
+            pattern = re.compile(compiled.pattern, flags | re.MULTILINE)
+        matches = list(pattern.finditer(text))
+        if invert:
+            if matches:
+                return "", {"mode": "multiline" if not dotall else "full", "total_lines": text.count("\n") + (1 if text else 0), "matched_lines": 0, "invert": True, "truncated": False, "byte_truncated": False}
+            output = text
+            meta = {"mode": "multiline" if not dotall else "full", "matched_lines": 1, "invert": True, "truncated": False, "byte_truncated": False}
+            lines = text.splitlines()
+            if max_lines > 0 and len(lines) > max_lines:
+                output = "\n".join(lines[:max_lines])
+                meta["truncated"] = True
+            meta["total_lines"] = len(lines)
+            meta["display_lines"] = min(len(lines), max_lines) if max_lines > 0 else len(lines)
+            return output, meta
+        chunks = []
+        for match in matches:
+            snippet = match.group(0)
+            if "\n" in snippet:
+                snippet = snippet.replace("\n", "\\n")
+            start_line = text.count("\n", 0, match.start()) + 1
+            chunks.append((start_line, f"  @{start_line}: {snippet}"))
+        entries, line_capped = cap_line_entries(chunks, max_lines if max_lines > 0 else len(chunks))
+        meta = {
+            "mode": "multiline" if not dotall else "full",
+            "total_lines": text.count("\n") + (1 if text else 0),
+            "matched_lines": len(matches),
+            "invert": False,
+            "truncated": line_capped,
+            "byte_truncated": False,
+            "display_lines": len(entries),
+        }
+        return "\n".join(line for _, line in entries), meta
+
+
+    def apply_body_limits_plain(text: str, max_lines: int) -> Tuple[str, Dict[str, object]]:
+        lines = text.splitlines()
+        meta = {
+            "mode": "plain",
+            "total_lines": len(lines),
+            "matched_lines": len(lines),
+            "invert": False,
+            "truncated": False,
+            "byte_truncated": False,
+            "display_lines": len(lines),
+        }
+        output = text
+        if max_lines > 0 and len(lines) > max_lines:
+            output = "\n".join(lines[:max_lines])
+            meta["truncated"] = True
+            meta["display_lines"] = max_lines
+        return output, meta
+
+
+    def format_response_body_output(
+        decoded_body: str,
+        filter_pattern: str,
+        filter_mode: str,
+        filter_invert: bool,
+        filter_ignore_case: bool,
+        max_lines: int,
+        max_bytes: int,
+        preview_lines: int,
+        context_lines: int,
+        compiled_filter=None,
+    ) -> Tuple[str, Dict[str, object]]:
+        text = decoded_body.rstrip("\r\n")
+        if not text:
+            return "", {"mode": "empty", "total_lines": 0, "matched_lines": 0, "invert": filter_invert, "truncated": False, "byte_truncated": False, "display_lines": 0}
+
+        lines = text.splitlines()
+        mode = (filter_mode or "line").strip().lower()
+        if mode not in {"line", "multiline", "full"}:
+            mode = "line"
+
+        if filter_pattern:
+            compiled = compiled_filter or compile_response_filter(filter_pattern, filter_ignore_case)
+            if mode == "line":
+                output, meta = filter_body_by_lines(lines, compiled, filter_invert, context_lines, max_lines)
+            else:
+                output, meta = filter_body_multiline(text, compiled, filter_invert, max_lines, dotall=(mode == "full"))
+            meta["filter_pattern"] = filter_pattern
+            if not output and not filter_invert:
+                preview = min(max(preview_lines, 0), len(lines))
+                if preview > 0:
+                    preview_text = format_line_entries(lines, list(range(1, preview + 1)), ellipsis_gaps=False)
+                    preview_text, byte_truncated = truncate_utf8(preview_text, max_bytes)
+                    return preview_text, {
+                        **meta,
+                        "preview": True,
+                        "matched_lines": 0,
+                        "display_lines": preview,
+                        "byte_truncated": byte_truncated,
+                    }
+                return "", {**meta, "preview": False, "matched_lines": 0, "display_lines": 0}
+        else:
+            output, meta = apply_body_limits_plain(text, max_lines)
+
+        output, byte_truncated = truncate_utf8(output, max_bytes)
+        if byte_truncated:
+            meta["byte_truncated"] = True
+        return output, meta
+
+
+    def print_response_body_summary(meta: Dict[str, object]):
+        mode = meta.get("mode")
+        if mode == "empty":
+            return
+        parts = [f"mode={mode}"]
+        if meta.get("filter_pattern"):
+            parts.append(f"pattern={meta['filter_pattern']!r}")
+        if meta.get("invert"):
+            parts.append("invert=true")
+        total = meta.get("total_lines")
+        matched = meta.get("matched_lines")
+        displayed = meta.get("display_lines")
+        if total is not None and matched is not None:
+            parts.append(f"matched {matched}/{total} lines")
+        if displayed is not None:
+            parts.append(f"showing {displayed}")
+        if meta.get("preview"):
+            parts.append("preview on zero match")
+        if meta.get("truncated"):
+            parts.append("line cap applied")
+        if meta.get("byte_truncated"):
+            parts.append("byte cap applied")
+        print(f"[body] {' | '.join(parts)}")
+
+
     def main():
         parser = argparse.ArgumentParser(description="Pure Python HTTP testing helper powered by httpx")
         parser.add_argument("--url", required=True)
@@ -466,6 +690,16 @@ args:
         parser.add_argument("--debug", dest="debug", action="store_true")
         parser.add_argument("--response-encoding", dest="response_encoding", default="")
         parser.add_argument("--download", dest="download", default="")
+        parser.add_argument("--response-filter", dest="response_filter", default="")
+        parser.add_argument("--response-filter-mode", dest="response_filter_mode", default="line")
+        parser.add_argument("--response-filter-invert", dest="response_filter_invert", action="store_true")
+        parser.add_argument("--no-response-filter-invert", dest="response_filter_invert", action="store_false")
+        parser.add_argument("--response-filter-ignore-case", dest="response_filter_ignore_case", action="store_true")
+        parser.add_argument("--no-response-filter-ignore-case", dest="response_filter_ignore_case", action="store_false")
+        parser.add_argument("--response-max-lines", dest="response_max_lines", type=int, default=0)
+        parser.add_argument("--response-max-bytes", dest="response_max_bytes", type=int, default=0)
+        parser.add_argument("--response-preview-lines", dest="response_preview_lines", type=int, default=5)
+        parser.add_argument("--response-context-lines", dest="response_context_lines", type=int, default=0)
         parser.set_defaults(
             include_headers=False,
             auto_encode_url=False,
@@ -475,9 +709,22 @@ args:
             show_command=False,
             show_summary=False,
             debug=False,
+            response_filter_invert=False,
+            response_filter_ignore_case=False,
         )
         args = parser.parse_args()
 
+        response_filter = (args.response_filter or "").strip()
+        response_max_lines = max(0, args.response_max_lines or 0)
+        response_max_bytes = max(0, args.response_max_bytes or 0)
+        response_preview_lines = max(0, args.response_preview_lines if args.response_preview_lines is not None else 5)
+        response_context_lines = max(0, args.response_context_lines or 0)
+        compiled_response_filter = None
+        if response_filter:
+            compiled_response_filter = compile_response_filter(
+                response_filter, args.response_filter_ignore_case
+            )
+
         repeat = max(1, args.repeat)
         try:
             delay_between = float(args.delay or "0")
@@ -648,9 +895,37 @@ args:
                             for key, value in response.headers.items():
                                 print(f"{key}: {value}")
                             print("")
-                        output_body = decoded_body.rstrip()
+                        output_body, body_output_meta = format_response_body_output(
+                            decoded_body,
+                            response_filter,
+                            args.response_filter_mode,
+                            args.response_filter_invert,
+                            args.response_filter_ignore_case,
+                            response_max_lines,
+                            response_max_bytes,
+                            response_preview_lines,
+                            response_context_lines,
+                            compiled_filter=compiled_response_filter,
+                        )
+                        has_filter_or_cap = bool(
+                            response_filter or response_max_lines > 0 or response_max_bytes > 0
+                        )
+                        if has_filter_or_cap and body_output_meta.get("mode") != "empty":
+                            print_response_body_summary(body_output_meta)
+                            if body_output_meta.get("preview") and not body_output_meta.get("matched_lines"):
+                                print("[body] no regex match; showing preview:")
                         if output_body:
                             print(output_body)
+                            if body_output_meta.get("truncated") or body_output_meta.get("byte_truncated"):
+                                omitted = (body_output_meta.get("total_lines") or 0) - (
+                                    body_output_meta.get("display_lines") or 0
+                                )
+                                if omitted > 0:
+                                    print(f"[body] ... {omitted} more line(s) omitted (use --download for full body)")
+                        elif body_output_meta.get("mode") == "empty":
+                            print("[no body]")
+                        elif response_filter and not body_output_meta.get("preview"):
+                            print("[body] no lines matched filter")
                         else:
                             print("[no body]")
 
@@ -729,6 +1004,13 @@ description: |
   - 连接探针：在无代理场景下额外进行 DNS/TCP/TLS 探测，粗粒度复刻 curl -w 指标
   - 可重复观测：repeat/delay + TTFB/total/speed_download 统计，便于盲注/时序测试
   - 扩展开关：additional_args 解析 http2/cert/verify/trust_env/max_redirects 等 httpx 选项
+  - 响应体瘦身：response_filter 按行/块正则提取，配合 max_lines/max_bytes 限制 stdout，降低 Agent token 消耗
+
+  **响应过滤最佳实践：**
+  - 大页面/HTML：用 `response_filter` 抓 error|exception|password|token|uid 等关键字行
+  - 无 filter 时：设 `response_max_lines=80` 或 `response_max_bytes=8192` 防止整页灌入上下文
+  - 0 命中：自动预览前 `response_preview_lines` 行，避免误判「空响应」
+  - 完整留存：大 body 用 `download` 落盘，stdout 只保留摘要行
 parameters:
   - name: "url"
     type: "string"
@@ -836,6 +1118,56 @@ parameters:
     description: "强制响应解码使用的编码（如GBK），覆盖自动探测"
     required: false
     flag: "--response-encoding"
+  - name: "response_filter"
+    type: "string"
+    description: |
+      响应体正则过滤（仅影响 stdout，不影响 --download 与指标）。
+      默认 line 模式按行匹配；示例：'(error|exception|SQL|password|token|uid)'。
+      与 response_max_lines/response_max_bytes 配合可显著减少 token 消耗。
+    required: false
+    flag: "--response-filter"
+  - name: "response_filter_mode"
+    type: "string"
+    description: "过滤模式：line（按行，默认）、multiline（跨行块）、full（整段 DOTALL 匹配）"
+    required: false
+    default: "line"
+    flag: "--response-filter-mode"
+  - name: "response_filter_invert"
+    type: "bool"
+    description: "反向过滤：输出不匹配 regex 的行（用于剔除 HTML 噪音）"
+    required: false
+    default: false
+    flag: "--response-filter-invert"
+  - name: "response_filter_ignore_case"
+    type: "bool"
+    description: "正则忽略大小写"
+    required: false
+    default: false
+    flag: "--response-filter-ignore-case"
+  - name: "response_max_lines"
+    type: "int"
+    description: "stdout 最多输出行数（0=不限制）；有 filter 时限制命中行数，无 filter 时截断全文"
+    required: false
+    default: 0
+    flag: "--response-max-lines"
+  - name: "response_max_bytes"
+    type: "int"
+    description: "stdout 响应体 UTF-8 字节上限（0=不限制），超出部分截断"
+    required: false
+    default: 0
+    flag: "--response-max-bytes"
+  - name: "response_preview_lines"
+    type: "int"
+    description: "filter 零命中时预览的前 N 行（默认 5，0=不预览）"
+    required: false
+    default: 5
+    flag: "--response-preview-lines"
+  - name: "response_context_lines"
+    type: "int"
+    description: "line 模式下命中行上下各保留 N 行上下文（类似 grep -C）"
+    required: false
+    default: 0
+    flag: "--response-context-lines"
   - name: "action"
     type: "string"
     description: "保留字段：标识调用意图（request, spider等），脚本内部不使用"