feat(sensor): add DeepSeek Harness parser

.github/workflows/sensor-dsh.yml

@@ -0,0 +1,23 @@
+name: DeepSeek Harness parser platforms
+on:
+  pull_request:
+    paths: ['Sensor/**', '.github/workflows/sensor-dsh.yml']
+  push:
+    branches: [main]
+    paths: ['Sensor/**', '.github/workflows/sensor-dsh.yml']
+permissions:
+  contents: read
+jobs:
+  parser:
+    strategy:
+      fail-fast: false
+      matrix:
+        os: [ubuntu-latest, macos-latest, windows-latest]
+    runs-on: ${{ matrix.os }}
+    steps:
+      - uses: actions/checkout@v7
+      - uses: astral-sh/setup-uv@v7
+      - run: uv sync --extra dev --python 3.12
+        working-directory: Sensor
+      - run: uv run pytest tests/test_dsh_parser.py -q
+        working-directory: Sensor

README.md

@@ -2,7 +2,7 @@
 
 [![uber/ADR | Trendshift](https://trendshift.io/api/badge/repositories/100696)](https://trendshift.io/repositories/100696?utm_source=repository-badge&utm_medium=badge&utm_campaign=badge-repository-100696)
 
-ADR (Agentic AI Detection and Response) is an enterprise security system for AI agents. It helps organizations secure employee-facing agents such as Cursor, Claude Code, Codex, and GitHub Copilot CLI, as well as customer-facing agents such as AI support agents.
+ADR (Agentic AI Detection and Response) is an enterprise security system for AI agents. It helps organizations secure employee-facing agents such as Cursor, Claude Code, Codex, GitHub Copilot CLI, and DeepSeek Harness, as well as customer-facing agents such as AI support agents.
 
 ADR is **deployed in production at Uber**, and the accompanying paper was accepted to **MLSys 2026**: [Paper PDF](docs/adr-paper.pdf) · [Slides PDF](docs/adr-mlsys-2026-slides.pdf)
 
@@ -23,7 +23,7 @@ This repository contains the open-source **ADR Discovery**, **ADR Sensor**, **AD
 | Path                                               | ADR component              | Description                                                                          |
 | -------------------------------------------------- | -------------------------- | ------------------------------------------------------------------------------------ |
 | [Discovery/](Discovery/)                           | ADR Discovery              | Inventory the AI apps, CLI agents, IDE extensions, model runtimes, and MCP servers on an endpoint, and flag unknown surfaces for review |
-| [Sensor/](Sensor/)                                 | ADR Observability          | Collect and normalize agent telemetry from Claude Code, Cursor, Codex, GitHub Copilot CLI, opencode, Claude Desktop, and others |
+| [Sensor/](Sensor/)                                 | ADR Observability          | Collect and normalize agent telemetry from Claude Code, Cursor, Codex, GitHub Copilot CLI, DeepSeek Harness, opencode, Claude Desktop, and others |
 | [Detection/](Detection/)                           | ADR Benchmark + Detection  | Dual-agent detector, 134 MCP servers, 304 benchmark tasks, baselines, figure scripts |
 | [docs/REPRODUCIBILITY.md](docs/REPRODUCIBILITY.md) | Evaluation                 | Step-by-step workflow to reproduce benchmark detection and paper figures             |
 

Sensor/README.md

@@ -18,6 +18,7 @@ ADR Sensor is a Python library that collects telemetry from AI coding agents to
 | **Claude Desktop**         | `claude_desktop` | JSONL audit logs                    | macOS, Windows         |
 | **OpenAI Codex CLI**       | `codex`          | JSONL + SQLite path catalogs        | macOS, Linux, Windows  |
 | **GitHub Copilot CLI**     | `copilot`        | JSONL (`~/.copilot/session-state/`) | macOS, Linux, Windows  |
+| **DeepSeek Harness**       | `dsh`            | JSONL/Zstandard (`~/.dsh/sessions/`) | macOS, Linux, Windows  |
 | **Warp Terminal**          | `warp`           | SQLite (`warp.sqlite`)              | macOS, Windows         |
 | **opencode**               | `opencode`       | SQLite (`opencode.db`) or JSON tree | macOS, Linux           |
 
@@ -91,14 +92,38 @@ MCP tools are namespaced by opencode as `<server>_<tool>`, so any tool that is n
 known built-in and contains an underscore is recorded as `tool_type: "mcp_tool"` with
 its `server_name` populated.
 
+### DeepSeek Harness
+
+The `dsh` source reads current DeepSeek Harness v3 session logs from
+`$DSH_HOME/sessions/`, or `~/.dsh/sessions/` when `DSH_HOME` is unset. DSH uses
+Zstandard-compressed logs by default; uncompressed JSONL v3 logs are also
+supported. When migrations leave several generations in one session directory,
+the parser follows DSH and considers only the highest generation. A session is
+ingested only when that generation is v3, so historical or future formats are
+not silently interpreted with the wrong schema.
+
+Tool arguments and results, PTC sub-dispatches, failure status, approvals,
+provider/model context, sandbox mode, permission preset, typed message content,
+and recorded token usage are normalized into the Sensor schema. Tool names do
+not carry a universal MCP server identity, so the parser does not guess one.
+The standard 14-day file lookback applies; use `--all-history` to include older
+sessions. Storage and event behavior were
+checked against DSH's pinned
+[JSONL persistence contract](https://github.com/deepseek-ai/deepseek-harness/blob/c291e7961a515f6d7af9304e7fd1d257929aef26/packages/session/session-persistence-jsonl/README.md)
+and [v3 event declarations](https://github.com/deepseek-ai/deepseek-harness/blob/c291e7961a515f6d7af9304e7fd1d257929aef26/packages/core/session/src/types.ts).
+
+```bash
+adr-sensor --source dsh
+```
+
 
 ## Architecture
 
 ```
 ┌─────────────────────────────────────────────────────────────────┐
 │                        AI Agent Logs                            │
 │         Claude, Cursor, Cline, Codex, Copilot CLI, Warp         │
-│                    Claude Desktop, opencode                     │
+│              Claude Desktop, opencode, DeepSeek Harness         │
 └───────────────────────────────┬─────────────────────────────────┘
                                 ▼
 ┌─────────────────────────────────────────────────────────────────┐
@@ -153,6 +178,7 @@ adr-sensor --source claude
 adr-sensor --source cursor
 adr-sensor --source codex
 adr-sensor --source copilot
+adr-sensor --source dsh
 adr-sensor --source claude_desktop
 adr-sensor --source opencode
 
@@ -363,7 +389,7 @@ builds its `--source` choices from `SOURCES`, so it picks the new agent up for f
 | --------------- | ------------------------------------------- |
 | Python          | 3.9, 3.10, 3.11, 3.12, 3.13                 |
 | Operating system| macOS, Linux, Windows                       |
-| Dependencies    | `tabulate` (runtime only — no native deps)  |
+| Dependencies    | `tabulate`, `zstandard`                     |
 
 Which sources yield data depends on the host OS and on which agents are installed;
 see the platform column in [Supported AI Agents](#supported-ai-agents). Sources that
@@ -375,6 +401,7 @@ cannot run on the current platform are skipped rather than failing.
 | ----------------- | -------------------------- | ----------------------------------------------------------------- |
 | `CODEX_HOME`      | Codex parser               | Codex data root containing `sessions/` and optional `state_*.sqlite` catalogs (default `~/.codex`) |
 | `COPILOT_HOME`    | Copilot parser             | Copilot CLI data root containing `session-state/` (default `~/.copilot`) |
+| `DSH_HOME`         | DeepSeek Harness parser    | Harness data root containing `sessions/` (default `~/.dsh`) |
 | `XDG_CACHE_HOME`  | `AgentObserver`            | Base for `--save-sessions` output (`$XDG_CACHE_HOME/adr_sensor`, default `~/.cache/adr_sensor`) |
 | `XDG_DATA_HOME`   | opencode parser            | Overrides the opencode data directory (default `~/.local/share/opencode`) |
 | `OPENCODE_DB`     | opencode parser            | Overrides the opencode SQLite filename or path (`:memory:` is ignored) |
@@ -427,6 +454,7 @@ adr-sensor/
 │   │   ├── claude_desktop_parser.py
 │   │   ├── codex_parser.py
 │   │   ├── copilot_parser.py
+│   │   ├── dsh_parser.py
 │   │   ├── opencode_parser.py
 │   │   └── warp_parser.py
 │   ├── schemas/

Sensor/adr_sensor/__init__.py

@@ -2,9 +2,9 @@
 ADR Sensor - Agentic Detection & Response
 
 Security observability library for AI coding agents. Collects telemetry from
-Claude Code, Cursor, Cline, OpenAI Codex CLI, GitHub Copilot CLI, Warp
-Terminal, opencode, and Claude Desktop Agent Mode (including Dispatch
-sessions) to enable threat detection and security monitoring.
+Claude Code, Cursor, Cline, OpenAI Codex CLI, GitHub Copilot CLI, DeepSeek
+Harness, Warp Terminal, opencode, and Claude Desktop Agent Mode (including
+Dispatch sessions) to enable threat detection and security monitoring.
 
 Usage:
     from adr_sensor import AgentObserver

Sensor/adr_sensor/cli.py

@@ -51,6 +51,7 @@ def main():
   adr-sensor --source cursor              Ingest Cursor IDE logs only
   adr-sensor --source claude_desktop      Ingest Claude Desktop agent-mode logs only (macOS/Windows)
   adr-sensor --source copilot             Ingest GitHub Copilot CLI logs only
+  adr-sensor --source dsh                 Ingest DeepSeek Harness logs only
   adr-sensor --source opencode            Ingest opencode logs only
   adr-sensor --save-sessions              Save individual session files
   adr-sensor --output-format jsonl        Export as JSONL

Sensor/adr_sensor/observer.py

@@ -28,6 +28,7 @@
 from .parsers.codex_parser import CodexParser
 from .parsers.copilot_parser import CopilotParser
 from .parsers.cursor_parser import CursorParser
+from .parsers.dsh_parser import DshParser
 from .parsers.opencode_parser import OpencodeParser
 from .parsers.warp_parser import WarpParser
 from .schemas.agent_event_schema import AgentEvent
@@ -60,6 +61,7 @@ class AgentObserver:
         ("warp", "Warp Terminal"),
         ("codex", "Codex"),
         ("copilot", "GitHub Copilot CLI"),
+        ("dsh", "DeepSeek Harness"),
         ("opencode", "opencode"),
     )
 
@@ -69,7 +71,7 @@ class AgentObserver:
         "claude_desktop": ("Darwin", "Windows"),
     }
 
-    CONTENT_AWARE_INCREMENTAL_SOURCES = frozenset({"codex", "copilot"})
+    CONTENT_AWARE_INCREMENTAL_SOURCES = frozenset({"codex", "copilot", "dsh"})
 
     def __init__(self, output_dir: Optional[Path] = None, max_age_days: Optional[int] = None):
         """Initialize the AgentObserver.
@@ -87,6 +89,7 @@ def __init__(self, output_dir: Optional[Path] = None, max_age_days: Optional[int
         self.copilot_parser = (
             CopilotParser(max_age_days=max_age_days) if max_age_days is not None else CopilotParser()
         )
+        self.dsh_parser = DshParser(max_age_days=max_age_days) if max_age_days is not None else DshParser()
         self.cline_parser = ClineParser(max_age_days=max_age_days) if max_age_days is not None else ClineParser()
         self.warp_parser = WarpParser(max_age_days=max_age_days) if max_age_days is not None else WarpParser()
         self.opencode_parser = (
@@ -129,7 +132,7 @@ def ingest_all(
 
         Args:
             source_filter: Which source to ingest. One of 'all', 'claude', 'cursor',
-                'claude_desktop', 'cline', 'warp', 'codex', 'copilot', 'opencode'.
+                'claude_desktop', 'cline', 'warp', 'codex', 'copilot', 'dsh', 'opencode'.
 
         Returns:
             Tuple of (agent_events, system_configs).

Sensor/adr_sensor/parsers/__init__.py

@@ -11,6 +11,7 @@
 from .codex_parser import CodexParser
 from .copilot_parser import CopilotParser
 from .cursor_parser import CursorParser
+from .dsh_parser import DshParser
 from .opencode_parser import OpencodeParser
 from .warp_parser import WarpParser
 
@@ -22,6 +23,7 @@
     "CodexParser",
     "CopilotParser",
     "CursorParser",
+    "DshParser",
     "OpencodeParser",
     "WarpParser",
 ]