fix: security hardening, executor no-ops, palinode start, 42 tests

Independent review by OpenAI Codex:
- Path traversal: null byte, symlink escape, commonpath checks (API + CLI)
- Executor: no-op detection, fact text normalization, stats only on real changes
- datetime.utcnow() → datetime.now(UTC) across all modules
- palinode start: rewritten with multiprocessing
- MCP: fixed elif/else syntax bug in palinode_lint handler
- Tests: 42 passing (symlink, null-byte, no-op, missing-field, CLI start)
- Docs: tool count and Python version aligned

Co-Authored-By: OpenAI Codex <noreply@openai.com>
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

Author: 3 people

README.md

@@ -24,7 +24,7 @@ If every service crashes, `cat` still works. No external database. No cloud depe
 
 Most agent memory systems are opaque databases you can't inspect, flat files that don't scale, or graph stores that require infrastructure. Palinode is **memory with provenance** — the only system where you can `git blame` every fact your agent knows.
 
-- **Git blame/diff/rollback as agent tools** — not just git-compatible files, but `palinode_diff`, `palinode_blame`, and `palinode_rollback` as first-class MCP tools your agent can call. [DiffMem](https://github.com/search?q=diffmem) and Git-Context-Controller are PoCs; Palinode ships 15 MCP tools including 5 git operations.
+- **Git blame/diff/rollback as agent tools** — not just git-compatible files, but `palinode_diff`, `palinode_blame`, and `palinode_rollback` as first-class MCP tools your agent can call. [DiffMem](https://github.com/search?q=diffmem) and Git-Context-Controller are PoCs; Palinode ships 17 MCP tools including 5 git operations.
 
 - **Operation-based compaction with a deterministic executor** — the LLM outputs structured ops (KEEP/UPDATE/MERGE/SUPERSEDE/ARCHIVE), a deterministic executor applies them. The LLM never touches your files directly. [All-Mem](https://arxiv.org/search/?query=all-mem+memory) does something similar on graph nodes; Palinode does it on plain markdown with git commits.
 
@@ -70,7 +70,7 @@ Most agent memory systems are opaque databases you can't inspect, flat files tha
 
 ### Integration ✅
 - **OpenClaw plugin** — lifecycle hooks for inject, extract, and capture
-- **MCP server** — 15 tools for Claude Code and any MCP client
+- **MCP server** — 17 tools for Claude Code and any MCP client
 - **FastAPI server** — HTTP API for programmatic access
 - **CLI** — command-line search, stats, reindex
 
@@ -132,7 +132,7 @@ graph TD
 
 ## Requirements
 
-- **Python 3.12+**
+- **Python 3.11+**
 - **Ollama** with `bge-m3` model (for embeddings — `ollama pull bge-m3`)
 - **Git** (for memory versioning)
 - A directory for your memory files (local, or a private git repo)
@@ -148,7 +148,7 @@ Optional:
 | Embeddings | BGE-M3 via Ollama |
 | Consolidation LLM | [OLMo 3.1 32B AWQ](https://huggingface.co/allenai/OLMo-3.1-32B-AWQ) via vLLM |
 | Hardware | RTX 5090 32GB (consolidation), any CPU (embeddings + API) |
-| Python | 3.12 |
+| Python | 3.11+ |
 | OS | Ubuntu 22.04 (Linux), macOS 14+ (development) |
 
 Other models should work — the consolidation prompt is model-agnostic. Smaller models (8B) may produce less reliable JSON for compaction operations; use `json-repair` (included) as a safety net.
@@ -531,7 +531,7 @@ Palinode is informed by research and ideas from several projects in the agent me
 
 ### Architecture Inspiration
 
-- **[LLM Knowledge Bases](https://gist.github.com/karpathy/442a6bf555914893e9891c11519de94f)** (Andrej Karpathy, April 2026) — The "compile, don't retrieve" pattern: LLM incrementally builds a structured markdown wiki from raw sources. Palinode implements this with git provenance, deterministic compaction, and 15 MCP tools.
+- **[LLM Knowledge Bases](https://gist.github.com/karpathy/442a6bf555914893e9891c11519de94f)** (Andrej Karpathy, April 2026) — The "compile, don't retrieve" pattern: LLM incrementally builds a structured markdown wiki from raw sources. Palinode implements this with git provenance, deterministic compaction, and 17 MCP tools.
 
 - **[OpenClaw](https://github.com/openclaw/openclaw)** — The plugin SDK, lifecycle hooks, and `MEMORY.md` pattern that Palinode extends and replaces. Palinode started as a better memory system for OpenClaw agents.
 

docs/INSTALL-CLAUDE-CODE.md

@@ -1,12 +1,12 @@
 # Installing Palinode with Claude Code
 
-Palinode gives Claude Code persistent memory via MCP — 14 tools for searching, saving, and managing memories across sessions. The `palinode-session` skill auto-captures milestones and decisions during coding, so your memory stays fresh without manual effort.
+Palinode gives Claude Code persistent memory via MCP — 17 tools for searching, saving, and managing memories across sessions. The `palinode-session` skill auto-captures milestones and decisions during coding, so your memory stays fresh without manual effort.
 
 ## Prerequisites
 
 - Claude Code installed (`npm install -g @anthropic-ai/claude-code`)
 - Palinode API running (see below)
-- Python 3.12+
+- Python 3.11+
 
 ---
 

palinode/api/server.py

@@ -16,7 +16,7 @@
 import httpx
 import subprocess
 import glob
-from datetime import datetime
+from datetime import UTC, datetime
 from typing import Any
 
 from fastapi import FastAPI, HTTPException
@@ -34,7 +34,7 @@ class JsonlFormatter(logging.Formatter):
     """Logging Formatter dictating a JSONL chronological schema format."""
     def format(self, record: logging.LogRecord) -> str:
         return json.dumps({
-            "timestamp": datetime.utcnow().isoformat() + "Z",
+            "timestamp": _utc_now().isoformat().replace("+00:00", "Z"),
             "level": record.levelname,
             "name": record.name,
             "message": record.getMessage()
@@ -54,6 +54,34 @@ def format(self, record: logging.LogRecord) -> str:
 
 # ── Auto-summary helpers ──────────────────────────────────────────────────────
 
+
+def _utc_now() -> datetime:
+    """Return a timezone-aware UTC timestamp."""
+    return datetime.now(UTC)
+
+
+def _memory_base_dir() -> str:
+    """Return the canonical memory root."""
+    return os.path.realpath(getattr(config, "memory_dir", config.palinode_dir))
+
+
+def _resolve_memory_path(file_path: str) -> tuple[str, str]:
+    """Resolve a relative memory path without allowing traversal outside memory_dir."""
+    if "\x00" in file_path:
+        raise HTTPException(status_code=400, detail="Null bytes are not allowed in paths")
+    if os.path.isabs(file_path):
+        raise HTTPException(status_code=403, detail="Absolute paths are not allowed")
+
+    base_dir = _memory_base_dir()
+    resolved = os.path.realpath(os.path.join(base_dir, file_path))
+    try:
+        within_root = os.path.commonpath([base_dir, resolved]) == base_dir
+    except ValueError as exc:
+        raise HTTPException(status_code=403, detail="Path traversal rejected") from exc
+    if not within_root:
+        raise HTTPException(status_code=403, detail="Path traversal rejected")
+    return base_dir, resolved
+
 def _generate_summary(content: str) -> str:
     """Invokes Ollama to produce a single-sentence logical summary of file memory.
 
@@ -163,12 +191,17 @@ def list_api(category: str | None = None, core_only: bool = False) -> list[dict[
     from palinode.core import parser
 
     results = []
-    base_dir = getattr(config, 'memory_dir', config.palinode_dir)
+    base_dir = _memory_base_dir()
     search_pattern = os.path.join(base_dir, "**/*.md")
 
     skip_dirs = {"daily", "archive", "inbox", "logs"}
 
     for filepath in glob.glob(search_pattern, recursive=True):
+        try:
+            if os.path.commonpath([base_dir, os.path.realpath(filepath)]) != base_dir:
+                continue
+        except ValueError:
+            continue
         rel_path = os.path.relpath(filepath, base_dir)
         parts = rel_path.split(os.sep)
 
@@ -207,19 +240,20 @@ def list_api(category: str | None = None, core_only: bool = False) -> list[dict[
 def read_api(file_path: str, meta: bool = False) -> dict[str, Any]:
     from palinode.core import parser
 
-    base_dir = getattr(config, 'memory_dir', config.palinode_dir)
-    
+    candidates = [file_path]
     if not file_path.endswith(".md"):
-        test_path = os.path.join(base_dir, file_path)
-        if not os.path.exists(test_path):
-             file_path += ".md"
-             
-    resolved = os.path.realpath(os.path.join(base_dir, file_path))
-    if not resolved.startswith(os.path.realpath(base_dir)):
-        raise HTTPException(status_code=403, detail="Path traversal rejected")
-        
-    if not os.path.exists(resolved):
-         raise HTTPException(status_code=404, detail="File not found")
+        candidates.append(f"{file_path}.md")
+
+    resolved = ""
+    for candidate in candidates:
+        _, resolved_candidate = _resolve_memory_path(candidate)
+        if os.path.exists(resolved_candidate):
+            file_path = candidate
+            resolved = resolved_candidate
+            break
+
+    if not resolved:
+        raise HTTPException(status_code=404, detail="File not found")
 
     try:
         with open(resolved, "r") as f:
@@ -617,11 +651,7 @@ def history_api(file_path: str, limit: int = 10) -> dict[str, Any]:
     Shows when and how a memory was created, updated, or superseded.
     """
     import subprocess
-    base_dir = os.path.abspath(getattr(config, 'memory_dir', config.palinode_dir))
-    full_path = os.path.abspath(os.path.join(base_dir, file_path))
-    
-    if not full_path.startswith(base_dir):
-        raise HTTPException(status_code=400, detail="Invalid path (path traversal detected)")
+    base_dir, full_path = _resolve_memory_path(file_path)
 
     if not os.path.exists(full_path):
         raise HTTPException(status_code=404, detail="File not found")

palinode/cli/__init__.py

@@ -1,5 +1,4 @@
 import click
-import asyncio
 from palinode.core.config import config
 from palinode.cli.search import search
 from palinode.cli.save import save
@@ -58,38 +57,38 @@ def main():
 @click.option("--api/--no-api", default=True, help="Run API server")
 def start(watcher, api):
     """Start Palinode services in the foreground."""
-    from palinode.api.server import APIServer
-    from palinode.indexer.sqlite import SQLiteIndexer
-    from palinode.ingest.watcher import Watcher
-    from rich.live import Live
-    from rich.panel import Panel
+    from multiprocessing import Process
     from rich.console import Console
+    from palinode.api.server import main as api_main
+    from palinode.indexer.watcher import main as watcher_main
 
     console = Console()
-    
-    async def run_services():
-        tasks = []
-        if api:
-            console.print("[green]Starting API server...[/green]")
-            api_service = APIServer(config)
-            tasks.append(api_service.start())
-            
-        if watcher:
-            console.print("[green]Starting watcher...[/green]")
-            indexer = SQLiteIndexer(config)
-            watcher_service = Watcher(config, indexer)
-            tasks.append(watcher_service.start())
-            
-        if not tasks:
-            console.print("[yellow]No services specified to start.[/yellow]")
-            return
-            
-        await asyncio.gather(*tasks)
+    processes = []
+
+    if api:
+        console.print("[green]Starting API server...[/green]")
+        processes.append(Process(target=api_main, daemon=False))
+
+    if watcher:
+        console.print("[green]Starting watcher...[/green]")
+        processes.append(Process(target=watcher_main, daemon=False))
+
+    if not processes:
+        console.print("[yellow]No services specified to start.[/yellow]")
+        return
 
     try:
-        asyncio.run(run_services())
+        for process in processes:
+            process.start()
+        for process in processes:
+            process.join()
     except KeyboardInterrupt:
         console.print("\n[yellow]Stopping services...[/yellow]")
+        for process in processes:
+            if process.is_alive():
+                process.terminate()
+        for process in processes:
+            process.join(timeout=5)
 
 @main.command()
 @click.option("--watcher/--no-watcher", default=True, help="Stop memory watcher")

palinode/cli/read.py

@@ -6,6 +6,24 @@
 from palinode.cli._format import print_result, get_default_format, OutputFormat
 
 
+def _resolve_memory_path(file_path: str) -> tuple[str, str]:
+    """Resolve a relative memory path without allowing traversal outside memory_dir."""
+    if "\x00" in file_path:
+        raise click.ClickException("Null bytes are not allowed in paths")
+    if os.path.isabs(file_path):
+        raise click.ClickException("Absolute paths are not allowed")
+
+    base_dir = os.path.realpath(config.memory_dir)
+    resolved = os.path.realpath(os.path.join(base_dir, file_path))
+    try:
+        within_root = os.path.commonpath([base_dir, resolved]) == base_dir
+    except ValueError as exc:
+        raise click.ClickException("Invalid path") from exc
+    if not within_root:
+        raise click.ClickException("Path traversal rejected")
+    return base_dir, resolved
+
+
 @click.command()
 @click.argument("file_path")
 @click.option("--format", "fmt", type=click.Choice(["text", "json"]), default=None, help="Output format")
@@ -21,21 +39,13 @@ def read(file_path, fmt, meta):
 
         palinode read projects/palinode-status.md --meta --format json
     """
-    # Resolve path relative to memory_dir
-    if os.path.isabs(file_path):
-        full_path = file_path
-    else:
-        full_path = os.path.join(config.memory_dir, file_path)
+    _, full_path = _resolve_memory_path(file_path)
 
     if not os.path.exists(full_path):
-        # Try with .md extension
-        if not full_path.endswith(".md"):
-            full_path_md = full_path + ".md"
-            if os.path.exists(full_path_md):
-                full_path = full_path_md
-            else:
-                raise click.ClickException(f"File not found: {file_path}")
-        else:
+        if not file_path.endswith(".md"):
+            file_path = f"{file_path}.md"
+            _, full_path = _resolve_memory_path(file_path)
+        if not os.path.exists(full_path):
             raise click.ClickException(f"File not found: {file_path}")
 
     with open(full_path, "r") as f: