feat(ava): subagent hardening with audit logging

Based on Anthropic's effective harnesses research.

New modules:
- Omni/Agent/AuditLog.hs: JSONL audit logging with SubagentId linking
- Omni/Agent/Tools/AvaLogs.hs: Tool for Ava to query her own logs
- Omni/Agent/Subagent/HARDENING.md: Design documentation

Key features:
- SubagentHandle with TVar status for async execution and polling
- spawnSubagentAsync, querySubagentStatus, waitSubagent, cancelSubagent
- User confirmation: spawn_subagent requires confirmed=true after approval
- Audit logs stored in $AVA_DATA_ROOT/logs/{ava,subagents}/
- CLI: ava logs [--last=N] [<subagent_id>]
- read_ava_logs tool for Ava self-diagnosis

Tasks: t-267, t-268, t-269, t-270, t-271

1 files changed, 17 insertions, 4 deletions

diff --git a/Omni/Agent/Paths.hs b/Omni/Agent/Paths.hs
index 6facdc6..6df6991 100644
--- a/Omni/Agent/Paths.hs
+++ b/Omni/Agent/Paths.hs
@@ -1,5 +1,5 @@
-{-# LANGUAGE NoImplicitPrelude #-}
 {-# LANGUAGE OverloadedStrings #-}
+{-# LANGUAGE NoImplicitPrelude #-}
 
 -- | Configurable paths for Ava data directories.
 --
@@ -11,6 +11,9 @@ module Omni.Agent.Paths
     outreachDir,
     userScratchRoot,
     userScratchDir,
+    logsDir,
+    avaLogsDir,
+    subagentLogsDir,
   )
 where
 
@@ -21,9 +24,10 @@ import System.FilePath ((</>))
 import System.IO.Unsafe (unsafePerformIO)
 
 avaDataRoot :: FilePath
-avaDataRoot = unsafePerformIO <| do
-  m <- lookupEnv "AVA_DATA_ROOT"
-  pure (fromMaybe "_/var/ava" m)
+avaDataRoot =
+  unsafePerformIO <| do
+    m <- lookupEnv "AVA_DATA_ROOT"
+    pure (fromMaybe "_/var/ava" m)
 {-# NOINLINE avaDataRoot #-}
 
 skillsDir :: FilePath
@@ -37,3 +41,12 @@ userScratchRoot = avaDataRoot </> "users"
 
 userScratchDir :: Text -> FilePath
 userScratchDir user = userScratchRoot </> Text.unpack user
+
+logsDir :: FilePath
+logsDir = avaDataRoot </> "logs"
+
+avaLogsDir :: FilePath
+avaLogsDir = logsDir </> "ava"
+
+subagentLogsDir :: FilePath
+subagentLogsDir = logsDir </> "subagents"