@@ -217,6 +217,7 @@ Full release notes with details on each version: [GitHub Releases](https://githu
 - Fix: an unresolved local JS/TS import (to a file absent from the scan) now emits a stable, portable `ref` target id instead of leaking a per-checkout absolute-path slug (#2457, thanks @rohit-jsfreaky).
 - Fix: `graphify benchmark` no longer crashes on a node whose label is `None` (#2674, thanks @Arthuro0103).
 - Feat: `graphify vibe install` adds full-parity Mistral Vibe (mistralai/mistral-vibe) support — skill file with `user-invocable: true` frontmatter (exposes `/graphify` as a native slash command in vibe's autocomplete), `AGENTS.md` always-on section, and two `pre_tool` hooks in `.vibe/hooks.toml` (matchers `grep` and `read_file` — vibe's snake-cased tool names per `vibe/core/tools/base.py:get_name`) that nudge toward `graphify query` for search/read operations. Global scope honors `VIBE_HOME`, hook ownership is `name`-based (never touches user-authored hooks that happen to shell out to graphify), and `hooks.toml` merges preserve hand-authored comments via `tomlkit`. Verified end-to-end against vibe 2.24.0.
+- Fix: `graphify vibe install` now also writes a writable `graphify-extract` subagent to `<scope>/agents/graphify-extract.toml` and the skill fragment dispatches Task calls with `subagent_type="graphify-extract"` (#2537 review, thanks @MusicalNinjaDad). Vibe's built-in `explore` subagent (`vibe/core/agents/models.py:EXPLORE`) is read-only — its `enabled_tools = ["grep", "read_file", "skill"]` cannot write chunk JSON — so semantic extraction inheriting it silently produced zero `.graphify_chunk_NN.json` files and the pipeline aborted. The new agent enables `write_file` scoped to `**/graphify-out/.graphify_chunk_*.json` only (no unbounded write grant), reuses the `explore` system prompt, and is registered/unregistered by `_install_vibe_extract_agent` / `_uninstall_vibe_extract_agent` symmetric to the hook flow. `uninstall_all` sweeps it alongside the rest.
 
 ## 0.9.40 (2026-08-11)
 
 
@@ -91,6 +91,10 @@
     _install_vibe_hook,
     _vibe_hooks_path,
     _vibe_hook_entries,
+    _install_vibe_extract_agent,
+    _uninstall_vibe_extract_agent,
+    _vibe_extract_agent_path,
+    _vibe_agents_dir,
     claude_install,
     claude_uninstall,
     codebuddy_install,
 
@@ -1815,23 +1815,115 @@ def _uninstall_vibe_agents_md(targets: list[Path]) -> None:
             print(f"AGENTS.md was empty after removal - deleted {target.resolve()}")
 
 
+_VIBE_EXTRACT_AGENT_NAME = "graphify-extract"
+
+# TOML body for the writable graphify-extract subagent. Vibe's AgentRegistry
+# scans `<scope>/agents/*.toml`; the file stem becomes the agent name
+# (`AgentProfile.from_toml` sets `name=path.stem`, see mistralai/mistral-vibe
+# vibe/core/agents/models.py), so this file MUST be named
+# `graphify-extract.toml`. Only the four whitelisted tools are enabled, and
+# `write_file` is scoped to graphify's own chunk output so nothing else in the
+# workspace is writable. The system prompt id `explore` keeps it lean — vibe's
+# read-only subagent uses the same prompt.
+_VIBE_EXTRACT_AGENT_TOML = """\
+# graphify-extract: writable subagent used by the graphify skill.
+# Managed by `graphify vibe install` -- do not hand-edit; changes are overwritten.
+# Named vs. the read-only `explore` builtin because Task-dispatched semantic
+# extraction has to write one chunk JSON per invocation, which `explore`
+# cannot do (its enabled_tools omit write_file).
+display_name = "Graphify Extract"
+description = "Writable subagent used by graphify's /graphify skill for semantic chunk extraction. Reads files, writes one graphify-out/.graphify_chunk_NN.json per Task, then exits."
+safety = "safe"
+agent_type = "subagent"
+system_prompt_id = "explore"
+enabled_tools = ["grep", "read_file", "write_file", "skill"]
+
+[tools.write_file]
+permission = "always"
+allowlist = ["**/graphify-out/.graphify_chunk_*.json"]
+
+[tools.read_file]
+permission = "always"
+
+[tools.grep]
+permission = "always"
+
+[tools.skill]
+permission = "always"
+"""
+
+
+def _vibe_agents_dir(project_dir: Path, *, project: bool) -> Path:
+    """Return the vibe agents dir for the requested scope (mirror hooks.toml layout)."""
+    if project:
+        return (project_dir or Path(".")) / ".vibe" / "agents"
+    return _vibe_home() / "agents"
+
+
+def _vibe_extract_agent_path(project_dir: Path, *, project: bool) -> Path:
+    """Full path of the graphify-extract.toml agent file for the requested scope."""
+    return _vibe_agents_dir(project_dir, project=project) / f"{_VIBE_EXTRACT_AGENT_NAME}.toml"
+
+
+def _install_vibe_extract_agent(project_dir: Path, *, project: bool) -> None:
+    """Write the writable graphify-extract subagent (idempotent).
+
+    Without this, the graphify skill dispatches Task subagents that inherit
+    vibe's read-only `explore` profile and cannot write chunk JSON files, so
+    Part B extraction silently produces zero chunks (see #2537 review).
+    """
+    target = _vibe_extract_agent_path(project_dir, project=project)
+    target.parent.mkdir(parents=True, exist_ok=True)
+    body = _VIBE_EXTRACT_AGENT_TOML
+    if target.exists() and target.read_text(encoding="utf-8") == body:
+        print(f"  {target}  ->  graphify-extract agent already registered (no change)")
+        return
+    if target.exists():
+        backup = target.with_name(target.name + ".graphify-bak")
+        shutil.copy2(target, backup)
+    target.write_text(body, encoding="utf-8")
+    print(f"  {target}  ->  graphify-extract subagent registered")
+
+
+def _uninstall_vibe_extract_agent(project_dir: Path, *, project: bool) -> None:
+    """Remove the graphify-extract.toml agent file (leave user-authored agents alone)."""
+    target = _vibe_extract_agent_path(project_dir, project=project)
+    if not target.exists():
+        return
+    target.unlink()
+    print(f"  {target}  ->  graphify-extract subagent removed")
+    # Prune the agents dir if empty and we own it (project scope only — the
+    # global VIBE_HOME/agents may hold unrelated user agents).
+    if project:
+        agents_dir = target.parent
+        try:
+            if agents_dir.is_dir() and not any(agents_dir.iterdir()):
+                agents_dir.rmdir()
+        except OSError:
+            pass
+
+
 def _print_vibe_install_summary(strict: bool) -> None:
     """Post-install summary + optional strict-mode note."""
     print()
     print("Mistral Vibe will now check the knowledge graph before answering")
     print("codebase questions and rebuild it after code changes.")
     print("Use /graphify in vibe to build or update the graph.")
+    print("Semantic extraction dispatches to the `graphify-extract` subagent")
+    print("(installed alongside the skill; vibe's read-only `explore` cannot")
+    print("write chunk files).")
     if strict:
         print("Strict mode: the first raw file read per session is blocked until")
         print("one `graphify query` runs (toggle with GRAPHIFY_HOOK_STRICT=0).")
 
 
 def _vibe_install(project_dir: Path | None = None, *, project: bool = False, strict: bool = False) -> None:
-    """Full-parity Vibe install: skill + AGENTS.md always-on + hooks.toml."""
+    """Full-parity Vibe install: skill + AGENTS.md always-on + hooks.toml + graphify-extract agent."""
     project_dir = project_dir or Path(".")
     skill_dst = _copy_skill_file("vibe", project=project, project_dir=project_dir)
     _install_vibe_agents_md(_vibe_agents_md_path(project_dir, project=project))
     _install_vibe_hook(project_dir, project=project, strict=strict)
+    _install_vibe_extract_agent(project_dir, project=project)
 
     if project:
         _print_project_git_add_hint([
@@ -1856,10 +1948,12 @@ def _vibe_uninstall(project_dir: Path | None = None, *, project: bool = False, r
     if project or (explicit_dir and not remove_user_skill):
         _remove_skill_file("vibe", project=True, project_dir=project_dir)
         _uninstall_vibe_hook(project_dir, project=True)
+        _uninstall_vibe_extract_agent(project_dir, project=True)
         agents_md_targets.append(_vibe_agents_md_path(project_dir, project=True))
     if remove_user_skill:
         _remove_skill_file("vibe", project=False)
         _uninstall_vibe_hook(project_dir, project=False)
+        _uninstall_vibe_extract_agent(project_dir, project=False)
         agents_md_targets.append(_vibe_agents_md_path(project_dir, project=False))
 
     _uninstall_vibe_agents_md(agents_md_targets)
 
@@ -261,11 +261,12 @@ Load files from `graphify-out/.graphify_uncached.txt`. Split into chunks of 20-2
 
 > Uses the `Task` tool for parallel subagent dispatch.
 > Call `Task` once per chunk — ALL in the same response so they run in parallel.
+> **On Mistral Vibe, dispatch each `Task` with `subagent_type="graphify-extract"`.** This writable subagent is installed by `graphify vibe install` alongside the skill. Do NOT use vibe's built-in `explore` subagent — it is read-only (`enabled_tools = ["grep", "read_file", "skill"]`) and cannot write the chunk JSON files this pipeline needs. If `graphify-extract` is missing (older install, or you overrode `VIBE_HOME` after install), re-run `graphify vibe install` first.
 
 Pass the extraction prompt as the task description:
 
 ```
-Task(description="Your task is to perform the following. Follow the instructions below exactly.\n\n<agent-instructions>\n[extraction prompt, with FILE_LIST, CHUNK_NUM, TOTAL_CHUNKS, DEEP_MODE substituted]\n</agent-instructions>\n\nExecute this now. Output ONLY the structured JSON response.")
+Task(subagent_type="graphify-extract", description="Your task is to perform the following. Follow the instructions below exactly.\n\n<agent-instructions>\n[extraction prompt, with FILE_LIST, CHUNK_NUM, TOTAL_CHUNKS, DEEP_MODE substituted]\n</agent-instructions>\n\nExecute this now. Output ONLY the structured JSON response.")
 ```
 
 Each subagent writes its result to its own `graphify-out/.graphify_chunk_NN.json`. Collect results as each `Task` completes and parse each as JSON.
@@ -285,10 +286,10 @@ See `references/extraction-spec.md` for the exact subagent prompt (JSON schema,
 Wait for all subagents. For each result:
 - Check that `graphify-out/.graphify_chunk_NN.json` exists on disk — this is the success signal
 - If the file exists and contains valid JSON with `nodes` and `edges`, include it and save to cache
-- If the file is missing, the subagent was likely dispatched as read-only (Explore type) — print a warning: "chunk N missing from disk — subagent may have been read-only. Re-run with general-purpose agent." Do not silently skip.
+- If the file is missing, the subagent was likely dispatched as read-only (vibe's built-in `explore`, or Claude Code's Explore type) — print a warning: "chunk N missing from disk — subagent may have been read-only. On vibe, re-dispatch with `subagent_type=\"graphify-extract\"` (installed by `graphify vibe install`). On Claude Code, re-dispatch with the general-purpose agent." Do not silently skip.
 - If a subagent failed or returned invalid JSON, print a warning and skip that chunk - do not abort
 
-If more than half the chunks failed or are missing, stop and tell the user to re-run and ensure `subagent_type="general-purpose"` is used.
+If more than half the chunks failed or are missing, stop and tell the user to re-run. On vibe, ensure `subagent_type="graphify-extract"` (installed by `graphify vibe install`) is passed to every `Task` call — the default `explore` subagent is read-only and will silently produce zero chunks. On Claude Code, ensure `subagent_type="general-purpose"` is used.
 
 Merge all chunk files into `.graphify_semantic_new.json`. **After each Agent call completes, read the real token counts from the Agent tool result's `usage` field and write them back into the chunk JSON before merging** — the chunk JSON itself always has placeholder zeros. Then run:
 ```bash