docs(#2628): update bash tool descriptions, docs, and system prompt for the todo model (trim impl details for agent)

Cargo.lock

@@ -1591,7 +1591,7 @@ name = "hive-bash-mcp"
 version = "0.1.0"
 dependencies = [
  "anyhow",
- "hive-core-agent-sock",
+ "hive-agent-sock",
  "hive-sh4re",
  "libc",
  "rmcp",

docs/persistence.md

@@ -233,13 +233,12 @@ Under `/var/lib/hyperhive/agents/<name>/`:
     (full captured output). `hive-c0re::bash_tasks_vacuum` runs
     hourly and deletes terminal task trios older than 48 hours;
     non-terminal (still-running) tasks are never deleted by vacuum.
-  - `mcp-loose-ends/` — JSON files published by external MCP daemons
-    (e.g. `hive-bash-mcp`, `hive-matrix-mcp`) listing their active
-    loose-end summary strings. Each file is `<daemon>.json` containing
-    a JSON array of plain-text lines. Read by `get_loose_ends` to
-    surface active background work without hardcoding per-MCP
-    knowledge in the harness. Files are created/removed by the
-    external daemons themselves.
+  - `hyperhive-todos.sqlite` — loose-ends-v2 todo store. In-container
+    MCP daemons (`hive-bash-mcp`, `hive-matrix-mcp`) and `forge_notify`
+    upsert keyed todos here over the harness's in-agent socket
+    (`HIVE_AGENT_SOCKET`); the harness merges them into `get_loose_ends`
+    output and clears a row on `mark_todo_done`. Replaced the old
+    file-based `mcp-loose-ends/` scanner.
   - `hyperhive-events.sqlite` — turn-loop event log.
   - `hyperhive-turn-stats.sqlite` — per-turn timing stats.
   - `hyperhive-model` — single-line model name override file.

docs/tools/bash.md

@@ -13,24 +13,24 @@ invocation regardless of tool groups.
 
 Submit a shell command for background execution (runs via `bash`).
 Stdout and stderr stream to `harness/bash-tasks/<id>.{out,err}`.
-When the task completes (or times out, or the process errors), the
-harness fires a wake with `from: "bash-task-<id>"`; the body contains
-the exit code and last stdout lines. Handle the completion on a future
-turn — unless `wait_seconds` already delivered the terminal result
-inline, in which case the wake is suppressed (see `status` below).
+When the task completes (or times out, or the process errors), it
+surfaces as a todo in the agent's loose-ends (via `get_loose_ends`),
+carrying the exit code and a `Read(<path>)` pointer to the captured
+output. Handle it on a future turn — unless `wait_seconds` already
+delivered the terminal result inline, in which case no todo is created
+(see `status` below).
 
 * `timeout_secs` — kill the task after N seconds and mark it
   `timed_out`. Omit for no timeout (runs until natural exit).
 * `wait_seconds` — inline poll before returning (capped at 30).
   When the task finishes within the window the full status is
-  returned immediately and no wake is fired; when the window expires
+  returned immediately and no todo is created; when the window expires
   the task keeps running and the normal `task started: id=<id>`
   response is returned. **Defaults to 3** — pass `wait_seconds: 0`
   to disable inline waiting and always get the immediate response.
 * `name` — optional caller-chosen task id. When set it replaces the
-  auto-generated hex id, so it surfaces in the wake `from`
-  (`bash-task-<name>`), in `status(<name>)` lookups, and in the
-  loose-ends list — a memorable label instead of an opaque id. A name
+  auto-generated hex id, so it surfaces in `status(<name>)` lookups and
+  the loose-ends list — a memorable label instead of an opaque id. A name
   is **reusable once its previous task has finished**; submitting a
   name whose task is still `pending`/`running` is rejected. Allowed
   characters: ASCII letters, digits, `.`, `_`, `-` (max 64). Omit for
@@ -52,15 +52,14 @@ finishes within the window the full status is returned immediately.
 Useful to avoid a separate round-trip when the task is expected to
 finish soon.
 
-Any `status` call (waited or not) that observes a terminal task
-suppresses that task's completion wake — you already have the result
-in this response, so no redundant `bash-task-<id>` inbox message
-follows (#2270). Narrow best-effort race: a `status`/`run` inline wait
-that resolves in the same instant the task actually finishes can still
-occasionally get both.
+Any `status` call (waited or not) that observes a terminal task clears
+that task's completion todo — you already have the result in this
+response, so no redundant loose-end follows. Narrow best-effort race: a
+`status`/`run` inline wait that resolves in the same instant the task
+actually finishes can still occasionally get both.
 
 Tasks marked `interrupted` had their process killed by a harness
-restart; a best-effort wake was still sent so the agent is not
+restart; a best-effort todo is still surfaced so the agent is not
 silently blocked.
 
 Exposed as `mcp__bash__status`.
@@ -68,8 +67,8 @@ Exposed as `mcp__bash__status`.
 ### `kill(id, force?)`
 
 Stop a running or pending task by its ID (from `run`). Fire-and-forget:
-sends the signal and returns without waiting — handle the completion
-wake (`from: "bash-task-<id>"`) on a future turn.
+sends the signal and returns without waiting — the completion surfaces
+in the agent's loose-ends; handle it on a future turn.
 
 - `force: false` (default) — SIGINT to the task's **process group**
   (graceful; lets the process clean up). The whole process group is
@@ -79,7 +78,7 @@ wake (`from: "bash-task-<id>"`) on a future turn.
 
 If a SIGINT'd task doesn't exit, call `kill` again with `force: true`.
 A still-pending task is cancelled before it starts. The task ends as
-`killed` and fires the usual completion wake.
+`killed` and surfaces in the loose-ends like any completion.
 
 Exposed as `mcp__bash__kill`.
 
@@ -97,7 +96,7 @@ matrix MCP:
 
 - **`hive-bash-daemon`** — long-running process (one per agent container,
   systemd service in `nix/agent-modules/mcp.nix`). Owns subprocess management,
-  output file writing, `mcp-loose-ends/` state, and wake signal delivery.
+  output file writing, and todo delivery on the harness's in-agent socket.
   Listens on `/run/hive-bash/socket` inside the container.
 
 - **`hive-bash-mcp`** — stdio bridge spawned by claude per turn (declared
@@ -108,13 +107,17 @@ matrix MCP:
 This split keeps claude's turn-local MCP bridge lightweight while the
 daemon tracks long-running tasks that outlive a single turn.
 
-### Transient wake (bypass broker sqlite)
+### Completion as a todo (loose-ends v2)
 
-When a bash task finishes, `hive-bash-daemon` sends the wake signal via
-the agent's per-agent socket as a **transient wake** (`AgentRequest::Wake`
-with `transient: true`). This bypasses broker sqlite for lower latency —
-the same mechanism used by matrix events. The message is delivered
-directly to the harness without touching the persistent message store.
+When a bash task changes state, `hive-bash-daemon` upserts a single keyed
+todo (`key = task id`) on the harness's in-agent socket (`HIVE_AGENT_SOCKET`)
+— "running" at start, then the completion summary when it finishes. The
+summary change signals the harness turn loop directly (in-process, no broker
+round-trip), so the agent is driven a turn to handle it via `get_loose_ends`,
+then clears the todo with `mark_todo_done`. Same mechanism the matrix daemon
+uses for unread rooms. An inline `wait_seconds` / `status` observation that
+already delivered the result instead clears the keyed todo, so no redundant
+loose-end follows.
 
 ## Relationship to the `execution` tool group
 

docs/tools/matrix.md

@@ -90,17 +90,16 @@ returned by `get_loose_ends` so unread rooms surface in the
 loose-ends list between turns.
 
 **Invite wakes**: when the daemon's sync loop receives an
-`m.room.member` invite event, it writes the invite to
-`mcp-loose-ends/matrix.json` and fires a hyperhive wake. The daemon
-does **not** auto-join — the agent calls `list_invites` to see pending
+`m.room.member` invite event, it upserts a todo (keyed `invite:<room>`)
+on the harness's in-agent socket, which drives a turn. The daemon does
+**not** auto-join — the agent calls `list_invites` to see pending
 invites and `resolve_invite` to accept or reject them.
 
-**Pending invites as loose ends**: pending invites are written to
-`mcp-loose-ends/matrix.json` and appear in `get_loose_ends` output as
+**Pending invites as loose ends**: pending invites are upserted as
+keyed todos and appear in `get_loose_ends` output as
 `[matrix] pending invite: <room> — use list_invites to see,
-resolve_invite to accept or reject`. The file is updated atomically
-after each invite event and after each `resolve_invite` (or
-`join_room`) call clears the entry.
+resolve_invite to accept or reject`. The keyed todo is cleared when a
+`resolve_invite` (or `join_room`) call resolves the invite.
 
 See [`docs/matrix.md`](../matrix.md) for the homeserver setup,
 provisioning flow, and federation config.

hive-agent-mcp/src/loose_ends.rs

@@ -1,42 +0,0 @@
-//! Generic scanner for MCP loose-end summary files.
-//!
-//! External MCP daemons (hive-bash-mcp, hive-matrix-mcp, etc.) write
-//! JSON files to `$HYPERHIVE_HARNESS_DIR/mcp-loose-ends/<name>.json`.
-//! Each file contains a JSON array of plain-text summary strings.
-//!
-//! The harness reads all files in this directory in `get_loose_ends` to
-//! surface active background work from any MCP without hardcoding
-//! per-MCP knowledge here.
-
-use std::path::PathBuf;
-
-/// Resolution lives in `hive_sh4re::paths` so the harness + every MCP
-/// daemon agree on where loose-end summary files are written.
-fn loose_ends_dir() -> PathBuf {
-    hive_sh4re::paths::mcp_loose_ends_dir()
-}
-
-/// Collect all loose-end summary strings published by external MCP daemons.
-/// Each string is a single line suitable for inclusion in `get_loose_ends`
-/// output. Returns an empty vec if the directory doesn't exist or is empty.
-#[must_use]
-pub fn collect() -> Vec<String> {
-    let Ok(rd) = std::fs::read_dir(loose_ends_dir()) else {
-        return Vec::new();
-    };
-    let mut out = Vec::new();
-    for entry in rd.flatten() {
-        let path = entry.path();
-        if path.extension().and_then(|e| e.to_str()) != Some("json") {
-            continue;
-        }
-        let Ok(content) = std::fs::read_to_string(&path) else {
-            continue;
-        };
-        let Ok(items) = serde_json::from_str::<Vec<String>>(&content) else {
-            continue;
-        };
-        out.extend(items);
-    }
-    out
-}

hive-agent-mcp/src/main.rs

@@ -18,7 +18,6 @@ use anyhow::Result;
 use clap::Parser;
 
 mod client;
-mod loose_ends;
 mod mcp;
 mod paths;
 mod send_allow;

hive-agent-mcp/src/mcp/mod.rs

@@ -339,20 +339,7 @@ impl AgentServer {
             if is_self_query && let Some(todos) = local_todos().await {
                 loose_ends.extend(todos);
             }
-            let mut out = annotate_retries(render_loose_ends(&loose_ends), retries);
-            // Append loose-end items published by external MCP daemons
-            // (e.g. active bash tasks from hive-bash-mcp). Generic — no
-            // per-MCP knowledge needed here.
-            let mcp_items = crate::loose_ends::collect();
-            if !mcp_items.is_empty() {
-                use std::fmt::Write as _;
-                let n = mcp_items.len();
-                let _ = write!(out, "\n\n{n} local task(s):");
-                for item in &mcp_items {
-                    let _ = write!(out, "\n- {item}");
-                }
-            }
-            out
+            annotate_retries(render_loose_ends(&loose_ends), retries)
         })
         .await
     }

hive-agent/prompts/system.md

@@ -8,7 +8,7 @@ Tools (hyperhive surface):
 - (some agents only) **extra MCP tools** surfaced as `mcp__<server>__<tool>` — these are agent-specific (matrix client, scraper, db connector, etc.) declared in your `agent.nix` under `hyperhive.extraMcpServers`. Treat them as first-class tools alongside the hyperhive surface; the operator already auto-approved them at deploy time.
 - `mcp__hyperhive__ask(question, options?, multi?, ttl_seconds?, to?)` — surface a structured question to the human operator (default, or `to: "operator"`) OR a peer agent (`to: "<agent-name>"`). Returns immediately with a question id — do NOT wait inline. When the recipient answers, a system message with event `question_answered { id, question, answer, answerer }` lands in your inbox; handle it on a future turn. Use this for clarifications, permission for risky actions, choice between options, or peer Q&A without burning regular inbox slots. `options` is advisory: a short fixed-choice list when applicable, otherwise leave empty for free text. `multi: true` lets the answerer pick multiple (checkboxes), answer comes back comma-joined. `ttl_seconds` auto-cancels with answer `[expired]` (and `answerer: "ttl-watchdog"`) when the decision becomes moot.
 - `mcp__hyperhive__answer(id, answer)` — answer a question that was routed to YOU. You'll see one in your inbox as a `question_asked { id, asker, question, options, multi }` system event when a peer or the operator calls `ask(to: "<your-name>", ...)`. The answer surfaces in the asker's inbox as a `question_answered` event. Strict authorisation: you can only answer questions where you are the declared target.
-- `mcp__hyperhive__get_loose_ends(agent?)` — list your loose ends: unanswered questions where you're asker (waiting on someone) or target (owing a reply), plus reminders you've scheduled that haven't fired. No args to list your own threads — cheap server-side sweep useful at turn start. Pass `agent: "<name>"` to inspect a peer agent's threads. Direct child agents are always accessible. For non-children, the `query_agent_state` capability is required — without it the request is rejected with an error.
+- `mcp__hyperhive__get_loose_ends(agent?)` — list your loose ends: unanswered questions where you're asker (waiting on someone) or target (owing a reply), reminders you've scheduled that haven't fired, plus in-container todos surfaced by your MCP daemons (finished/running background bash tasks, unread matrix rooms, forge threads). No args to list your own threads — cheap server-side sweep useful at turn start. Pass `agent: "<name>"` to inspect a peer agent's threads. Direct child agents are always accessible. For non-children, the `query_agent_state` capability is required — without it the request is rejected with an error.
 - `mcp__hyperhive__cancel_loose_end(kind, id)` — cancel one of your own open threads. `kind` is `"question"` (the asker — you, in this case — gets a `[cancelled by <you>]` answer so the waiter unblocks), `"reminder"` (hard-deleted before it fires), or `"approval"` (withdraws a pending approval you submitted that got superseded — root agent only; the server rejects this kind for all other callers). `id` from the matching `get_loose_ends` row or the original submission reply.
 - `mcp__hyperhive__remind(message, delay_seconds? | at_unix_timestamp?, file_path?)` — schedule a message to land in your _own_ inbox at a future time (sender shows as `reminder`). Set exactly one of `delay_seconds` (relative) or `at_unix_timestamp` (absolute). Use for self-paced follow-ups instead of blocking a whole turn on a long `recv` wait. A large `message` auto-spills to a file under `/agents/{label}/state/reminders/`; pass `file_path` to point at one yourself. Each agent's pending-reminder count is capped (default 50) — the tool will error if the cap is already reached.
 - `mcp__hyperhive__set_status(text)` — set a free-text status visible on the operator dashboard. **Call this at the start of every task** to say what you're working on (e.g. `"processing matrix messages"`, `"fixing #319 model priority"`, `"idle"`). Single line, ≤200 chars — the dashboard renders this as a short chip, so longer multi-line text is rejected. Pass an empty string to clear. Persists across harness restarts.

hive-agent/src/forge_notify.rs

@@ -64,6 +64,8 @@ const NEW_ITEM_TOLERANCE_SECS: i64 = 120;
 /// configured. Otherwise loops forever, polling every
 /// `POLL_INTERVAL_SECS` seconds. Errors are never fatal.
 ///
+/// `socket` is the harness's in-agent todo socket (`HIVE_AGENT_SOCKET`):
+/// each forge notification is pushed as an `upsert_todo`, not a direct wake.
 pub async fn run(socket: PathBuf) {
     let forge_url = match std::env::var("HIVE_FORGE_URL") {
         Ok(u) if !u.is_empty() => u,
@@ -1064,33 +1066,40 @@ async fn poll_once(
             continue;
         };
 
-        let req = hive_core_agent_sock::Request::Wake {
-            from: "forge".to_owned(),
-            body,
+        // Upsert a *todo* (loose-ends v2) on the harness's in-agent socket,
+        // keyed by the forge thread id, instead of firing a direct wake. A
+        // new/changed summary makes the harness signal its turn loop; the
+        // agent clears the todo (`mark_todo_done`) once it has handled the
+        // thread. Re-scanning the same thread is an idempotent no-op.
+        let req = hive_agent_sock::Request::UpsertTodo {
+            subsystem: "forge".to_owned(),
+            key: Some(id.to_string()),
+            summary: body,
+            source: None,
         };
-        let deliver_result =
-            crate::client::request::<_, hive_core_agent_sock::Response>(socket, &req)
-                .await
-                .map(|_| ());
+        let deliver_result = crate::client::request::<_, hive_agent_sock::Response>(socket, &req)
+            .await
+            .map(|_| ());
         match deliver_result {
             Ok(()) => {
-                debug!(%id, "forge_notify: delivered");
-                // Mark the thread read on forge immediately after a
-                // successful broker delivery. The broker inbox is the
-                // durable work queue now (each row has its own ack
-                // lifecycle), so the forge unread flag no longer needs to
-                // track agent processing — clearing it on delivery keeps
-                // forge's unread set tiny by construction, so a container
-                // rebuild re-scan finds nothing stale to re-deliver. The
-                // in-memory `delivered` entry below is only a within-process
-                // guard so a transient mark-read failure doesn't re-fire the
-                // wake next tick; it is deliberately NOT persisted — forge's
-                // own read-state is the cross-rebuild source of truth.
+                debug!(%id, "forge_notify: todo upserted");
+                // Mark the thread read on forge immediately after upserting
+                // the todo. The todo is the durable work item now (it stays
+                // in `get_loose_ends` until the agent marks it done), so the
+                // forge unread flag no longer needs to track agent
+                // processing — clearing it keeps forge's unread set tiny by
+                // construction, so a container rebuild re-scan finds nothing
+                // stale (and idempotent re-upserts wouldn't re-wake anyway).
+                // The in-memory `delivered` entry below is only a
+                // within-process guard so a transient mark-read failure
+                // doesn't re-upsert next tick; it is deliberately NOT
+                // persisted — forge's own read-state is the cross-rebuild
+                // source of truth.
                 mark_read(forge, id).await;
                 delivered.insert(id, updated_at);
             }
             Err(e) => {
-                warn!(%id, error = ?e, "forge_notify: deliver failed — leaving unread");
+                warn!(%id, error = ?e, "forge_notify: todo upsert failed — leaving unread");
             }
         }
     }

hive-agent/src/main.rs

@@ -421,7 +421,15 @@ async fn serve_main<S: Surface>(socket: &Path, poll_ms: u64) -> Result<()> {
     for failure in plugins::install_configured().await {
         S::send_to_parent(socket, failure).await;
     }
-    tokio::spawn(crate::forge_notify::run(socket.to_path_buf()));
+    // forge_notify pushes forge notifications as todos on the in-agent
+    // socket (loose-ends v2), not direct wakes — so it dials
+    // `HIVE_AGENT_SOCKET`, not the host mcp.sock.
+    tokio::spawn(crate::forge_notify::run(
+        std::env::var_os("HIVE_AGENT_SOCKET").map_or_else(
+            || std::path::PathBuf::from(hive_agent_sock::DEFAULT_AGENT_SOCKET),
+            std::path::PathBuf::from,
+        ),
+    ));
     // Agent-side cleanup of this agent's own harness artifacts (completed
     // bash-task files + verbose event rows). Runs here, not host-side in
     // hive-c0re, because the files are agent-owned — see `vacuum` module docs.

hive-bash-mcp/Cargo.toml

@@ -8,7 +8,7 @@ workspace = true
 
 [dependencies]
 anyhow.workspace = true
-hive-core-agent-sock.workspace = true
+hive-agent-sock.workspace = true
 hive-sh4re.workspace = true
 libc.workspace = true
 rmcp.workspace = true

hive-bash-mcp/src/bin/mcp.rs

@@ -125,7 +125,7 @@ fn render_bash_run(id: &str, resp: Result<DaemonResponse>) -> String {
 /// immediately re-waiting on the same task.
 const BASH_IDLE_WAIT_HINT: &str = "\n\nThe task is still running — your wait timed out before it \
 finished. If you have other useful work, do that and check back later (the task keeps running, and \
-a wake fires when it completes) rather than immediately re-waiting.";
+it surfaces in your loose-ends when it completes) rather than immediately re-waiting.";
 
 /// Turn a `DaemonResponse` from a `BashStatus` call into the string
 /// claude sees as the tool result. When `waited` is set (the call
@@ -154,7 +154,7 @@ fn render_bash_kill(resp: Result<DaemonResponse>) -> String {
                 let sig = payload["signal"].as_str().unwrap_or("SIGINT");
                 format!(
                     "task `{id}`: {sig} sent to its process group; it transitions to `killed` \
-                     once the process exits and the usual completion wake fires."
+                     once the process exits and then surfaces in your loose-ends."
                 )
             } else {
                 format!("task `{id}` was pending — cancelled before it started.")
@@ -180,17 +180,16 @@ struct BashRunArgs {
     timeout_secs: Option<u64>,
     /// Optional inline wait: `run` polls for up to `wait_seconds`
     /// (capped at 30) before returning. When the task finishes within the
-    /// window the full status is returned immediately and no wake is fired;
-    /// when the timeout expires the task keeps running and the normal
-    /// `task started: id=<id>` response is returned. Defaults to 3s. Pass
-    /// `0` to disable inline waiting and always get the immediate response.
+    /// window the full status is returned immediately (nothing to handle
+    /// later); when the timeout expires the task keeps running and the
+    /// normal `task started: id=<id>` response is returned. Defaults to 3s.
+    /// Pass `0` to disable inline waiting and always get the immediate response.
     #[serde(default = "default_wait")]
     wait_seconds: Option<u64>,
     /// Optional task name. When set it becomes the task id, so it appears
-    /// in the completion wake (`from: "bash-task-<name>"`), in `status`
-    /// lookups, and in the loose-ends list — handy for recognising a task
-    /// later instead of an opaque hex id. A name can be reused once its
-    /// previous task has finished; reusing a name whose task is still
+    /// in `status` lookups and your loose-ends list — handy for recognising
+    /// a task later instead of an opaque hex id. A name can be reused once
+    /// its previous task has finished; reusing a name whose task is still
     /// running is rejected. Allowed chars: ASCII letters, digits, `.`,
     /// `_`, `-` (max 64). Omit to get the auto-generated id.
     #[serde(default)]
@@ -237,21 +236,18 @@ struct BashMcp;
 impl BashMcp {
     #[tool(
         description = "Run a shell command in the background. Returns a task ID immediately — \
-            do NOT wait inline. When the command finishes, the harness fires a wake with \
-            `from: \"bash-task-<id>\"` carrying the exit status plus a `Read(<path>)` \
-            pointer to the captured `.out`/`.err` files (not the output text itself — \
-            read only what you need); handle it on a future turn. Use `status` to poll the task status within \
-            the same turn if needed. `timeout_secs` defaults to `None` (no timeout) — \
-            task runs until natural exit; pass an explicit value to kill after N seconds. \
-            Pass `wait_seconds` (capped at 30) to wait inline for fast commands: when the \
-            task finishes within the window the full status is returned immediately and no \
-            wake is fired; when the timeout expires the task keeps running and the normal \
-            `task started: id=<id>` response is returned. `wait_seconds` defaults to 3; \
-            pass `wait_seconds: 0` to disable inline waiting and always get the immediate \
-            response. Pass `name` to label the task with a memorable id (used in the wake \
-            `from`, `status` lookups, and the loose-ends list) instead of an opaque hex id; \
-            a name is reusable once its prior task has finished, and rejected while one is \
-            still running."
+            do NOT wait inline. When the command finishes it surfaces as a todo in your \
+            loose-ends (the exit status plus a `Read(<path>)` pointer to the captured \
+            `.out`/`.err` files — read only what you need); handle it on a future turn. Use \
+            `status` to poll within the same turn if needed. `timeout_secs` defaults to \
+            `None` (no timeout) — task runs until natural exit; pass a value to kill after \
+            N seconds. Pass `wait_seconds` (capped at 30) to wait inline for fast commands: \
+            if the task finishes within the window you get the full status immediately \
+            (nothing to handle later); otherwise it keeps running and you get \
+            `task started: id=<id>`. Defaults to 3; pass `0` to disable inline waiting. \
+            Pass `name` to label the task with a memorable id (shown in `status` lookups \
+            and your loose-ends) instead of an opaque hex id; reusable once the prior task \
+            has finished, rejected while one is still running."
     )]
     async fn run(&self, Parameters(args): Parameters<BashRunArgs>) -> String {
         let req = DaemonRequest::BashRun {
@@ -297,8 +293,8 @@ impl BashMcp {
             signalled, so a runaway child (cargo/nix/etc.) is stopped too, not just the shell. \
             Fire-and-forget: sends the signal and returns without waiting. If a SIGINT'd task \
             doesn't exit, call kill again with `force: true` to SIGKILL. A still-pending task \
-            is cancelled before it starts. The task ends as `killed` and fires the usual \
-            completion wake."
+            is cancelled before it starts. The task ends as `killed` and surfaces in your \
+            loose-ends like any completion."
     )]
     async fn kill(&self, Parameters(args): Parameters<BashKillArgs>) -> String {
         let req = DaemonRequest::BashKill {

hive-bash-mcp/src/main.rs

@@ -1,7 +1,8 @@
 //! `hive-bash-daemon` binary — long-running per-agent bash task runner.
 //! Spawns `sh -c` subprocesses, monitors completion, writes task state
-//! files, and fires hyperhive wake signals. Listens on a unix socket
-//! for tool-call requests from the `hive-bash-mcp` stdio bridge.
+//! files, and surfaces task state to the agent as todos on the harness's
+//! in-agent socket. Listens on a unix socket for tool-call requests from
+//! the `hive-bash-mcp` stdio bridge.
 
 use anyhow::Result;
 
@@ -15,17 +16,17 @@ async fn main() -> Result<()> {
         .init();
 
     let socket_path = hive_bash_mcp::paths::daemon_socket();
-    let wake_socket = hive_bash_mcp::paths::hyperhive_socket();
+    let todo_socket = hive_bash_mcp::paths::agent_socket();
 
     tracing::info!(
         socket = %socket_path.display(),
-        wake = %wake_socket.display(),
+        todo = %todo_socket.display(),
         "hive-bash-daemon starting"
     );
 
     // Start the background runner loop — scans for pending tasks and
-    // spawns them, sends wake signals on completion.
-    hive_bash_mcp::runner::spawn_loop(wake_socket);
+    // spawns them, pushing todos to the harness on task transitions.
+    hive_bash_mcp::runner::spawn_loop(todo_socket);
 
     // Serve the unix socket forever.
     hive_bash_mcp::socket::serve(&socket_path).await

hive-bash-mcp/src/paths.rs

@@ -43,21 +43,17 @@ pub fn turn_stats_db() -> PathBuf {
     harness_dir().join("hyperhive-turn-stats.sqlite")
 }
 
-/// Hyperhive control socket — the daemon writes wake signals here so
-/// the harness drives a new claude turn on bash task completion.
-/// Mirrors the path used by `forge_notify` and `hive-matrix-mcp`.
+/// The harness-served in-agent todo socket (loose-ends v2). The runner
+/// pushes bash-task todos here — `upsert_todo` while a task is active,
+/// a keyless `done` todo on completion — instead of firing a direct
+/// c0re wake. Override via `HIVE_AGENT_SOCKET`; mirrors the path the
+/// harness binds (`agent-service.nix`) and the matrix producer dials.
 #[must_use]
-pub fn hyperhive_socket() -> PathBuf {
-    std::env::var_os("HIVE_CONTROL_SOCKET")
-        .map_or_else(|| PathBuf::from("/run/hive/mcp.sock"), PathBuf::from)
-}
-
-/// Directory where MCP daemons write loose-end summary files for the harness.
-/// Each daemon writes `<name>.json` here; the harness scans the dir in
-/// `get_loose_ends` to surface active work from all MCPs generically.
-#[must_use]
-pub fn mcp_loose_ends_dir() -> PathBuf {
-    hive_sh4re::paths::mcp_loose_ends_dir()
+pub fn agent_socket() -> PathBuf {
+    std::env::var_os("HIVE_AGENT_SOCKET").map_or_else(
+        || PathBuf::from(hive_agent_sock::DEFAULT_AGENT_SOCKET),
+        PathBuf::from,
+    )
 }
 
 /// Full path for a task's JSON metadata file.

hive-bash-mcp/src/runner.rs

@@ -1,15 +1,22 @@
 //! Bash subprocess runner: spawns `bash -c <cmd>` tasks, writes status
-//! files under `harness/bash-tasks/`, and fires hyperhive wake signals
-//! on completion.
+//! files under `harness/bash-tasks/`, and surfaces task state to the agent
+//! as *todos* on the harness's in-agent socket (loose-ends v2).
 //!
 //! Files under `tasks_dir()`:
 //! - `<id>.json`  — task metadata + status (pending → running → done)
 //! - `<id>.out`   — captured stdout (streamed while running)
 //! - `<id>.err`   — captured stderr (streamed while running)
 //!
+//! Todos (loose-ends v2): one keyed todo (`key = task id`) tracks the task
+//! across its lifetime — upserted with a "running" summary at start, then
+//! upserted again with the completion summary when it finishes. That summary
+//! change signals the harness turn loop the same way the old completion wake
+//! did; the agent clears the todo with `mark_todo_done` once it has read the
+//! output.
+//!
 //! Tasks with status `running` on daemon boot are marked `interrupted`
-//! (the process died with the previous daemon). A best-effort wake is
-//! still sent so the agent is not silently blocked.
+//! (the process died with the previous daemon). The todo is still updated to
+//! its interrupted-done summary so the agent is not silently blocked.
 //!
 //! The runner kills the child process on timeout — `tokio::process::Child::drop()`
 //! does not kill children, so we explicitly call `child.kill().await`.
@@ -21,6 +28,7 @@ use std::sync::{Arc, Mutex, OnceLock};
 use std::time::{Duration, SystemTime, UNIX_EPOCH};
 
 use anyhow::{Result, bail};
+use hive_agent_sock::Request as TodoReq;
 use tokio::io::AsyncWriteExt as _;
 use tokio::sync::Notify;
 
@@ -157,41 +165,84 @@ pub fn read_task(id: &str) -> Option<TaskFile> {
 }
 
 // ---------------------------------------------------------------------------
-// Loose-ends file (generic MCP loose-ends protocol)
+// Todo delivery (loose-ends v2)
 // ---------------------------------------------------------------------------
+//
+// Bash-task state is surfaced to the agent as todos on the harness's
+// in-agent socket (`HIVE_AGENT_SOCKET`), not as direct c0re wakes: one keyed
+// todo (`key = task id`) tracks the task across its whole lifetime. It's
+// upserted with a stable "running" summary when the task starts (an
+// idempotent no-op that never re-wakes), then upserted again with the
+// completion summary when it finishes — that summary change signals the
+// harness turn loop exactly like the old completion wake did. The agent
+// reads the output and clears the todo with `mark_todo_done`.
 
-/// Rewrite `mcp-loose-ends/bash.json` with a summary of all currently
-/// active (Pending or Running) tasks. The harness scans this directory
-/// generically in `get_loose_ends` — no bash-specific code needed there.
-///
-/// File format: a JSON array of plain-text summary strings, one per
-/// loose-end item. The harness includes them verbatim in the output.
-/// Atomic write (tmp + rename) so the harness never reads a partial file.
-fn refresh_loose_ends() {
-    let active = active_tasks();
-    let dir = paths::mcp_loose_ends_dir();
-    if let Err(e) = std::fs::create_dir_all(&dir) {
-        tracing::warn!(error = ?e, "bash_runner: create mcp-loose-ends dir failed");
-        return;
-    }
-    let items: Vec<String> = active
-        .iter()
-        .map(|t| {
-            let age = now_unix() - t.created_at;
-            format!(
-                "bash task `{}` status={:?}, cmd: `{}`, age {}s",
-                t.id, t.status, t.cmd, age
-            )
-        })
-        .collect();
-    let dest = dir.join("bash.json");
-    let tmp = dest.with_extension("json.tmp");
-    let json = serde_json::to_string(&items).unwrap_or_else(|_| "[]".to_owned());
-    if let Err(e) = std::fs::write(&tmp, &json).and_then(|()| std::fs::rename(&tmp, &dest)) {
-        tracing::warn!(error = ?e, "bash_runner: write mcp-loose-ends/bash.json failed");
+/// Send one todo request to the harness in-agent socket. Best-effort: any
+/// connect / write error is logged and swallowed — the harness self-heals
+/// on the next task transition, and a standalone daemon without the socket
+/// simply has nowhere to push.
+async fn send_todo(socket: &Path, req: &TodoReq) {
+    use tokio::io::{AsyncBufReadExt as _, BufReader};
+    use tokio::net::UnixStream;
+    let line = match serde_json::to_string(req) {
+        Ok(mut s) => {
+            s.push('\n');
+            s
+        }
+        Err(e) => {
+            tracing::warn!(error = ?e, "bash_runner: serialise todo failed");
+            return;
+        }
+    };
+    match UnixStream::connect(socket).await {
+        Ok(stream) => {
+            let (read, mut write) = stream.into_split();
+            if write.write_all(line.as_bytes()).await.is_err() {
+                tracing::warn!("bash_runner: write todo failed");
+                return;
+            }
+            let _ = write.shutdown().await;
+            // Drain the response so the server doesn't get ECONNRESET.
+            let mut resp = String::new();
+            let _ = BufReader::new(read).read_line(&mut resp).await;
+        }
+        Err(e) => {
+            tracing::warn!(error = ?e, socket = %socket.display(), "bash_runner: connect todo socket failed");
+        }
     }
 }
 
+/// Upsert the task's keyed todo (`key = id`) with `summary`. Used for both
+/// the "running" surface at start and the "done" summary at completion — a
+/// changed summary on the same row signals the turn loop, an unchanged
+/// re-push is an idempotent no-op.
+async fn upsert_bash_todo(socket: &Path, id: &str, summary: String) {
+    send_todo(
+        socket,
+        &TodoReq::UpsertTodo {
+            subsystem: "bash".to_owned(),
+            key: Some(id.to_owned()),
+            summary,
+            source: None,
+        },
+    )
+    .await;
+}
+
+/// Clear a task's keyed todo — used when an inline `status`/`run` wait
+/// already delivered the terminal result, so no `done` todo is warranted.
+async fn clear_bash_todo(socket: &Path, id: &str) {
+    send_todo(
+        socket,
+        &TodoReq::ClearTodo {
+            subsystem: "bash".to_owned(),
+            key: Some(id.to_owned()),
+            all: false,
+        },
+    )
+    .await;
+}
+
 // ---------------------------------------------------------------------------
 // Public API used by daemon dispatch
 // ---------------------------------------------------------------------------
@@ -265,7 +316,10 @@ pub fn submit_task(cmd: String, timeout_secs: Option<u64>, name: Option<String>)
         stderr_tail: None,
     };
     write_task(&task)?;
-    refresh_loose_ends();
+    // No todo yet: the keyed "active" todo is upserted when the runner
+    // loop flips the task to Running (`run_task`), which has the in-agent
+    // socket handle. Surfacing a Pending task would only race the ~200ms
+    // until it starts and double-wake (pending→running).
     Ok(id)
 }
 
@@ -385,7 +439,9 @@ pub fn kill_task(id: &str, force: bool) -> (bool, bool) {
         task.status = TaskStatus::Killed;
         task.completed_at = Some(now_unix());
         let _ = write_task(&task);
-        refresh_loose_ends();
+        // A Pending task never reached Running, so it has no keyed "active"
+        // todo to clear and (like the old code) fires no completion signal —
+        // the caller that killed it already knows.
         return (true, false);
     }
     (false, false)
@@ -395,9 +451,9 @@ pub fn kill_task(id: &str, force: bool) -> (bool, bool) {
 // Runner background loop
 // ---------------------------------------------------------------------------
 
-/// Spawn the background runner loop as a detached tokio task. `socket`
-/// is the path to the per-agent broker socket used to deliver completion
-/// wake signals. Call once at daemon startup.
+/// Spawn the background runner loop as a detached tokio task. `socket` is
+/// the harness's in-agent todo socket (`HIVE_AGENT_SOCKET`), where task
+/// transitions are pushed as todos. Call once at daemon startup.
 pub fn spawn_loop(socket: PathBuf) {
     tokio::spawn(async move {
         run_loop(socket).await;
@@ -443,8 +499,12 @@ async fn mark_interrupted(socket: &Path) {
         if let Err(e) = write_task(&task) {
             tracing::warn!(id = %id, error = ?e, "bash_runner: write interrupted state failed");
         }
-        refresh_loose_ends();
-        send_wake(socket, &id, "interrupted (daemon restarted)", None).await;
+        upsert_bash_todo(
+            socket,
+            &id,
+            done_summary(&id, "interrupted (daemon restarted)", None),
+        )
+        .await;
     }
 }
 
@@ -494,7 +554,12 @@ async fn run_task(mut task: TaskFile, socket: &Path) {
     if let Err(e) = write_task(&task) {
         tracing::warn!(id = %id, error = ?e, "bash_runner: write running state failed");
     }
-    refresh_loose_ends();
+    upsert_bash_todo(
+        socket,
+        &id,
+        format!("bash task `{id}` running: `{}`", task.cmd),
+    )
+    .await;
     // Best-effort: tally the normalised command head for the /stats
     // "favorite tools" view. Counted once per execution, regardless of
     // exit status. Never fails the task.
@@ -562,25 +627,33 @@ async fn run_task(mut task: TaskFile, socket: &Path) {
     if let Err(e) = write_task(&task) {
         tracing::warn!(id = %id, error = ?e, "bash_runner: write done state failed");
     }
-    refresh_loose_ends();
-
-    // Skip the wake if a `status`/`run` inline wait already handed this
-    // exact terminal result to the caller in a tool response — see
-    // `wait_for_task` / `observe_terminal`. Narrow race: an inline waiter
-    // that polls in the few hundred ms right around this point may lose the
-    // race and still get a wake alongside its inline result; best-effort,
-    // same tolerance as the rest of this daemon's delivery guarantees.
+    // If an inline `status`/`run` wait already handed this exact terminal
+    // result to the caller in a tool response (see `wait_for_task` /
+    // `observe_terminal`), there's nothing left to surface — retire the
+    // keyed todo without a `done` upsert so the agent gets no redundant
+    // wake. Narrow race: an inline waiter polling in the few hundred ms
+    // around this point may lose the race and still get a `done` todo
+    // alongside its inline result; best-effort, same tolerance as the rest
+    // of this daemon's guarantees.
     if take_wake_suppressed(&id) {
-        tracing::debug!(id = %id, "bash_runner: wake suppressed (already observed via status)");
+        clear_bash_todo(socket, &id).await;
+        tracing::debug!(id = %id, "bash_runner: done todo suppressed (already observed via status)");
         return;
     }
 
-    // Pass only whether each stream produced (trimmed) output — `send_wake`
-    // emits a `Read(<path>)` pointer to the captured `.out`/`.err` files
-    // rather than inlining the tail into the wake body.
+    // Transition the SAME keyed todo from "running" to its done summary: the
+    // summary change signals the turn loop, and the agent clears the todo
+    // with `mark_todo_done` after reading the output. Pass only whether each
+    // stream produced (trimmed) output — the summary carries a `Read(<path>)`
+    // pointer to the captured `.out`/`.err` files rather than inlining the tail.
     let has_stdout = !stdout_tail.as_deref().unwrap_or("").trim().is_empty();
     let has_stderr = !stderr_tail.as_deref().unwrap_or("").trim().is_empty();
-    send_wake(socket, &id, &summary, Some((has_stdout, has_stderr))).await;
+    upsert_bash_todo(
+        socket,
+        &id,
+        done_summary(&id, &summary, Some((has_stdout, has_stderr))),
+    )
+    .await;
 }
 
 /// Run `bash -c cmd` in its own process group, streaming output to files.
@@ -726,24 +799,20 @@ fn tail_file(path: &Path, max_bytes: usize) -> Option<String> {
 }
 
 // ---------------------------------------------------------------------------
-// Wake delivery
+// Completion summary
 // ---------------------------------------------------------------------------
 
-pub(crate) async fn send_wake(
-    socket: &Path,
-    id: &str,
-    summary: &str,
-    // `(has_stdout, has_stderr)` — whether the task produced non-empty
-    // stdout / stderr. `None` for the interrupted path (no captured output
-    // to point at). Deliberately NOT the output text: the wake gives the
-    // agent the status + a `Read(<path>)` pointer to the captured files
-    // rather than inlining the tail, which would flood the agent's context
-    // on every task completion.
-    output: Option<(bool, bool)>,
-) {
+/// Build the completion-summary line carried by a `done` todo: the status
+/// `summary` plus `Read(<path>)` pointers to any captured output.
+///
+/// `output` is `(has_stdout, has_stderr)` — whether the task produced
+/// non-empty stdout / stderr; `None` for the interrupted path, which has no
+/// captured output. Deliberately NOT the output text: the todo gives the
+/// agent the status + a `Read(<path>)` pointer to the captured files rather
+/// than inlining the tail, which would flood the agent's context on every
+/// task completion.
+fn done_summary(id: &str, summary: &str, output: Option<(bool, bool)>) -> String {
     use std::fmt::Write as _;
-    use tokio::io::{AsyncBufReadExt as _, BufReader};
-    use tokio::net::UnixStream;
     let mut body = format!("bash task `{id}` finished: {summary}");
     if let Some((has_stdout, has_stderr)) = output
         && (has_stdout || has_stderr)
@@ -760,38 +829,7 @@ pub(crate) async fn send_wake(
             );
         }
     }
-    let req = hive_core_agent_sock::Request::Wake {
-        from: format!("bash-task-{id}"),
-        body,
-    };
-
-    match UnixStream::connect(socket).await {
-        Ok(stream) => {
-            let (read, mut write) = stream.into_split();
-            let line = match serde_json::to_string(&req) {
-                Ok(mut s) => {
-                    s.push('\n');
-                    s
-                }
-                Err(e) => {
-                    tracing::warn!(id = %id, error = ?e, "bash_runner: serialise wake failed");
-                    return;
-                }
-            };
-            if write.write_all(line.as_bytes()).await.is_err() {
-                tracing::warn!(id = %id, "bash_runner: write wake failed");
-                return;
-            }
-            let _ = write.shutdown().await;
-            // Drain the response so the server doesn't get ECONNRESET.
-            let mut resp = String::new();
-            let _ = BufReader::new(read).read_line(&mut resp).await;
-            tracing::info!(id = %id, "bash_runner: wake delivered");
-        }
-        Err(e) => {
-            tracing::warn!(id = %id, error = ?e, "bash_runner: connect wake socket failed");
-        }
-    }
+    body
 }
 
 #[cfg(test)]

hive-sh4re/src/paths.rs

@@ -28,10 +28,3 @@ pub fn harness_dir() -> PathBuf {
     let label = std::env::var("HIVE_LABEL").unwrap_or_default();
     PathBuf::from(format!("/agents/{label}/harness"))
 }
-
-/// Directory where out-of-process MCP daemons write loose-end summary
-/// files (`<name>.json`) for the harness to scan in `get_loose_ends`.
-#[must_use]
-pub fn mcp_loose_ends_dir() -> PathBuf {
-    harness_dir().join("mcp-loose-ends")
-}

nix/agent-modules/mcp.nix

@@ -175,7 +175,11 @@ in
       ];
       environment = {
         HIVE_BASH_SOCKET = "/run/hive-bash/socket";
-        HIVE_CONTROL_SOCKET = "/run/hive/mcp.sock";
+        # In-agent todo socket the harness serves (loose-ends v2): the
+        # runner pushes bash-task todos here (upsert while active, keyless
+        # 'done' on completion) instead of firing a c0re wake. Must match
+        # the harness's HIVE_AGENT_SOCKET (agent-service.nix).
+        HIVE_AGENT_SOCKET = "/run/hive-agent/${userName}/agent.sock";
         RUST_LOG = "info";
         # HYPERHIVE_HARNESS_DIR and HYPERHIVE_STATE_DIR are already
         # injected via systemd.globalEnvironment by the meta flake