Merge pull request #318 from andrewyng/feat/code-agent-upgrades

Code agent: tool ergonomics, parallel execution, plan/discuss modes, explorer subagent

Author: drp8226

platform/coworker/agent.py

@@ -19,6 +19,7 @@
     make_send_message_tool,
 )
 from .engine import Approver, TurnEngine
+from .environment import environment_context
 from .memory import MemoryStore, Scope, format_memories, memory_tools
 from .permissions import Mode, PermissionEngine
 from .project import load_agents_md
@@ -28,10 +29,43 @@
 from .skills import SkillLoader, skill_catalog_text, skill_tools
 from .tools import ToolRegistry
 from .tools.directories import request_directory_tool
+from .tools.plan import propose_plan_tool
+from .tools.subagent import explorer_tools
 from .web import make_web_fetch_tool, make_web_search_tool
 from .tools.shell import LocalExecutor
 from .tools.todo import TodoList
 
+# Appended each turn while discuss mode is active: enforcement-only read-only, with no
+# pressure toward a plan proposal (that's what distinguishes it from plan mode).
+_DISCUSS_MODE_CONTEXT = """\
+Discuss mode is active: write and shell tools are disabled. Explore and answer freely; if
+the user asks for a change, describe it in chat instead of attempting it (they can switch
+to plan or approval mode to have you make it)."""
+
+# Appended to the latest user message every turn while plan mode is active. The mode can
+# flip mid-session (plan approval), so this can't live in the static instructions.
+_PLAN_MODE_CONTEXT = """\
+Plan mode is active: write and shell tools are blocked. Explore read-only and design an
+approach. When you've committed to one, present it with `propose_plan` (what you'll change,
+in which files, how you'll verify) — don't describe edits as if you were making them. If
+the plan is approved, this same session switches to execution and you implement it; if
+rejected, revise the plan using the feedback."""
+
+# When-to-remember rules, injected only when a memory store is wired. Without these,
+# models either never call `remember` or save noise the repo already records.
+_MEMORY_GUIDANCE = """\
+Memory:
+- You have persistent memory across sessions. Use `remember` for durable facts: the user's \
+corrections and stated preferences (include the why), and project context you couldn't \
+rederive from the code. Don't save what the repo already records (code structure, git \
+history, AGENTS.md) or details that only matter to the current task. Use absolute dates, \
+never "yesterday".
+- Before saving, check the known-memories list: if an entry already covers it, revise that \
+entry with `memory_update` instead of adding a near-duplicate; retire wrong or obsolete \
+entries with `memory_forget`.
+- Memories reflect when they were written. If one names a file, flag, or URL, verify it \
+still exists before relying on it."""
+
 
 def _enabled_connector_tools(secrets: SecretStore) -> tuple[set[str], set[str]]:
     connectors = {c["name"]: c for c in connector_list(secrets)}
@@ -77,6 +111,7 @@ def build_engine(
     audit_sink: Optional[Any] = None,
     roots: Optional[list] = None,
     directory_requester: Optional[Any] = None,
+    plan_approver: Optional[Any] = None,
 ) -> TurnEngine:
     ws = Path(workspace).expanduser().resolve() if workspace else None
     if agent.needs_workspace and ws is None:
@@ -129,6 +164,21 @@ def build_engine(
     # Web search + fetch: research tools for every agent (keyless DuckDuckGo default).
     registry.register(make_web_search_tool(secrets))
     registry.register(make_web_fetch_tool())
+    # Route by the model's `provider:` prefix (OpenAI default, Ollama, …). The manager normally
+    # passes its shared router; this fallback covers the TUI / direct build_engine() callers.
+    # Resolved here (not at engine construction) because the explorer subagent captures it.
+    provider = provider or ProviderRouter(secrets, default_provider="openai")
+    # The Code agent can fan broad research out to read-only explorer subagents, keeping its
+    # own context for the actual change.
+    if agent.name == "code" and ws is not None:
+        registry.register_all(
+            explorer_tools(
+                workspace=ws,
+                provider=provider,
+                model=model,
+                model_settings=model_settings,
+            )
+        )
     # Scheduling: Cowork + MyHelper can set up scheduled tasks (origin = this session).
     if (
         task_store is not None
@@ -147,6 +197,7 @@ def build_engine(
 
     instructions = agent.system_prompt
     if ws is not None:
+        instructions = f"{instructions}\n\n{environment_context(ws)}"
         conventions = load_agents_md(ws)
         if conventions:
             instructions = f"{instructions}\n\n{conventions}"
@@ -155,6 +206,7 @@ def build_engine(
         registry.register_all(
             memory_tools(memory_store, workspace=str(ws) if ws else None)
         )
+        instructions = f"{instructions}\n\n{_MEMORY_GUIDANCE}"
         remembered = memory_store.list(scope=Scope.GLOBAL)
         if ws is not None:
             remembered += memory_store.list(scope=Scope.WORKSPACE, workspace=str(ws))
@@ -175,17 +227,32 @@ def build_engine(
         auto_allow_tools=set(config.auto_allow),
         roots=root_list or None,
     )
-    # Tell the agent, each turn, which directories it has and their access (orphan Cowork can gain
-    # folders mid-session) — appended to the latest user message since mid-thread system messages
-    # aren't reliable across providers. Multi-dir surfaces (Cowork/MyHelper) only.
-    context_provider = (
+    # The plan-mode exit door. Always registered (surfaces can flip a live session into
+    # plan mode via set_mode, and the registry is fixed at build); the engine rejects the
+    # call whenever the session isn't actually in plan mode.
+    registry.register(propose_plan_tool())
+
+    # Per-turn ephemeral context, appended to the latest user message since mid-thread system
+    # messages aren't reliable across providers. Two producers: the plan-mode reminder (mode can
+    # flip mid-session, so it's checked each turn, not baked into the instructions) and the live
+    # directory list (orphan Cowork can gain folders mid-session; Cowork/MyHelper only).
+    roots_context = (
         (lambda: render_context(root_list))
         if root_list and agent.name in ("cowork", "myhelper")
         else None
     )
-    # Route by the model's `provider:` prefix (OpenAI default, Ollama, …). The manager normally
-    # passes its shared router; this fallback covers the TUI / direct build_engine() callers.
-    provider = provider or ProviderRouter(secrets, default_provider="openai")
+
+    def context_provider() -> str:
+        parts = []
+        if permissions.mode is Mode.PLAN:
+            parts.append(_PLAN_MODE_CONTEXT)
+        elif permissions.mode is Mode.DISCUSS:
+            parts.append(_DISCUSS_MODE_CONTEXT)
+        if roots_context is not None:
+            ctx = roots_context()
+            if ctx:
+                parts.append(ctx)
+        return "\n\n".join(parts)
 
     engine = TurnEngine(
         provider=provider,
@@ -202,6 +269,7 @@ def build_engine(
         audit_sink=audit_sink,
         context_provider=context_provider,
         directory_requester=directory_requester,
+        plan_approver=plan_approver,
     )
     engine.executor = executor  # type: ignore[attr-defined]
     engine.todo = todo  # type: ignore[attr-defined]

platform/coworker/agents/code.py

@@ -4,6 +4,7 @@
 
 import aisuite as ai
 
+from ..tools.files import file_tools
 from ..tools.git import git_tools
 from ..tools.search import search_tools
 from ..tools.shell import shell_tools
@@ -17,6 +18,12 @@
 - Explore first. Use `grep` and `read_file` to find the relevant code and learn how it works \
 before editing. Don't guess at APIs, signatures, or layout — read them. `git_log` shows how a \
 file evolved. Read meaningful chunks, not a line at a time.
+- Independent lookups run in parallel: when you need several reads/greps and none depends on \
+another's result, request them together in one batch instead of one per turn.
+- For broad questions spanning many files ("where is X handled?", "how does the Y flow \
+work?"), delegate to `explore` — a read-only subagent that searches in its own context and \
+returns only a report, keeping your context for the actual change. Independent explores can \
+run in parallel. For a single known file, just read it yourself.
 
 Match the codebase:
 - Write code that reads like the surrounding code: match its style, naming, structure, and \
@@ -35,10 +42,13 @@
 `apply_unified_diff` for standard unified diffs; `write_file` for new files or full rewrites.
 
 Verify:
-- `run_shell` is a persistent bash (cd and env persist). After changes, run the narrowest \
+- `run_shell` is a persistent shell (cd and env persist). After changes, run the narrowest \
 relevant test/build/lint to confirm your work. Don't report something done without verifying \
 it; if you can't verify, say so plainly. Don't repeat a failing command — if stuck after 2–3 \
 attempts, step back, reconsider, and surface the blocker.
+- Pass a short `description` with each command (shown in approval prompts), and raise \
+`timeout_seconds` for slow builds/tests. For long-running processes (dev servers, watchers), \
+set `run_in_background` and poll `shell_task_output`; stop them with `shell_task_kill`.
 
 Plan multi-step work:
 - For anything beyond a few steps, maintain a task list with `todo_write`: keep exactly one \
@@ -59,14 +69,17 @@
 def code_agent() -> Agent:
     def factory(context: AgentContext) -> list:
         ws = str(context.workspace)
-        # Our `grep` (ripgrep, .gitignore-aware) replaces the toolkit's slower search_files.
+        # Our `grep` (ripgrep, .gitignore-aware) replaces the toolkit's slower search_files;
+        # our line-numbered, windowing `read_file` replaces its read_file/read_file_lines.
+        replaced = {"search_files", "read_file", "read_file_lines"}
         files = [
             t
             for t in ai.toolkits.files(root=ws, allow_write=True)
-            if getattr(t, "__name__", "") != "search_files"
+            if getattr(t, "__name__", "") not in replaced
         ]
         tools = [
             *files,
+            *file_tools(ws),  # read_file (numbered lines, windowed)
             *ai.toolkits.git(root=ws),  # git_status, git_diff
             *git_tools(ws),  # git_log
             *search_tools(ws),  # grep