claude-recall 0.20.10 → 0.20.12

package/.claude/settings.json

@@ -1,5 +1,27 @@
 {
   "hooks": {
+    "SubagentStart": [
+      {
+        "hooks": [
+          {
+            "type": "command",
+            "command": "node /home/ebiarao/.nvm/versions/node/v20.19.3/lib/node_modules/claude-recall/dist/cli/claude-recall-cli.js hook run subagent-start",
+            "timeout": 5
+          }
+        ]
+      }
+    ],
+    "SubagentStop": [
+      {
+        "hooks": [
+          {
+            "type": "command",
+            "command": "node /home/ebiarao/.nvm/versions/node/v20.19.3/lib/node_modules/claude-recall/dist/cli/claude-recall-cli.js hook run subagent-stop",
+            "timeout": 10
+          }
+        ]
+      }
+    ],
     "SessionStart": [
       {
         "matcher": "compact",
@@ -88,5 +110,5 @@
       }
     ]
   },
-  "hooksVersion": "11.0.0"
+  "hooksVersion": "12.0.0"
 }

package/.claude/skills/auto-corrections/SKILL.md

@@ -8,28 +8,17 @@ source: claude-recall
 
 # Corrections
 
-Auto-generated from 18 memories. Last updated: 2026-04-06.
+Auto-generated from 7 memories. Last updated: 2026-04-06.
 
 ## Rules
 
 - CORRECTION: Memory with complex metadata
+- CORRECTION: Never delete claude-recall memories without asking the user first. Show what you plan to delete and get explicit approval before running delete_memory.
 - CORRECTION: Memory with complex metadata
 - CORRECTION: Memory with complex metadata
-- CORRECTION: Memory with complex metadata
-- CORRECTION: Memory with complex metadata
-- CORRECTION: Memory with complex metadata
-- CORRECTION: Memory with complex metadata
-- CORRECTION: Memory with complex metadata
-- CORRECTION: Memory with complex metadata
-- CORRECTION: Memory with complex metadata
-- CORRECTION: Memory with complex metadata
-- CORRECTION: Memory with complex metadata
-- CORRECTION: Memory with complex metadata
-- CORRECTION: Memory with complex metadata
-- CORRECTION: Memory with complex metadata
-- CORRECTION: License copyright should include user's name instead of 'Claude Recall Contributors'
+- CORRECTION: i just want you tellme what you observe claude recall has done lately
 - CORRECTION: Use true agentic design where agents decide when to run, not cron jobs
-- CORRECTION: License copyright should list your name instead of 'Claude Recall Contributors'
+- CORRECTION: delete without asking first
 
 ---
 *Auto-generated by Claude Recall. Regenerate: `npx claude-recall skills generate`*

package/.claude/skills/auto-corrections/manifest.json

@@ -1,26 +1,15 @@
 {
   "topicId": "corrections",
-  "sourceHash": "f96988b031c72141d7607ba19dc871b09a774b4cebd97a35cb4155ac8b1f4055",
-  "memoryCount": 18,
-  "generatedAt": "2026-04-06T18:15:28.374Z",
+  "sourceHash": "c3bf72dbbe482ded411952a50d3c0d6bb04f7c0a8d8ee76c01992ed8f27c8fcd",
+  "memoryCount": 7,
+  "generatedAt": "2026-04-06T21:44:03.092Z",
   "memoryKeys": [
+    "memory_1775511843079_fzzjymikj",
+    "memory_1775511566528_2zqlbeiyy",
+    "memory_1775500740471_dp5hmdxaf",
     "memory_1775499328353_2ade6746o",
-    "memory_1775496241485_1yyc5ht88",
-    "memory_1775494279035_j6uj5lzxo",
-    "memory_1775492069326_vksvzmt3f",
-    "memory_1775491767369_sepsjmg8y",
-    "memory_1775169786543_43p8to1hu",
-    "memory_1775169704632_wzwczltzu",
-    "memory_1775169639101_rmxkftqtk",
-    "memory_1775169273281_4fod7ku6f",
-    "memory_1775169256118_y7kc07y6y",
-    "memory_1775169244673_d6d0uc33a",
-    "memory_1775160781806_vnqf355e9",
-    "memory_1775159399542_awfryrxp3",
-    "memory_1775157082825_8uvrul28i",
-    "memory_1775156381094_dr1nihes8",
-    "hook_correction_1774180798682_9b5a2hadq",
+    "hook_correction_1775511596163_oj20xa10q",
     "hook_correction_1775498361284_9amjuw4le",
-    "hook_correction_1774180805192_4ft4zhfsa"
+    "hook_correction_1775511572913_yq1y9tvlo"
   ]
 }

package/.claude/skills/auto-failure-lessons/SKILL.md

@@ -8,16 +8,41 @@ source: claude-recall
 
 # Failure Lessons
 
-Auto-generated from 27 memories. Last updated: 2026-04-06.
+Auto-generated from 87 memories. Last updated: 2026-04-06.
 
 ## Rules
 
 - Check command syntax, file paths, and prerequisites before running
+- Check inputs and prerequisites before retrying
+- OpenClaw gateway websocket connection fails from root user context in sandbox; this is a known pairing regression. CLI cannot communicate with gateway from inside the container.
 - SQLite query syntax error: LIKE clause requires single quotes around string literal, not double quotes
+- Fault injection into mock-eiap/world-state.json on host will not be visible to sandbox agents until collect.sh runs and pushes updated signals.json into the sandbox. Verify data propagation by checking signals.json inside sandbox, not just host files.
+- Session extraction prompt design flaw: fix pairing relies on string similarity between commands, but problem fixes often look structurally different from the failing command (e.g. bash script vs echo piped to bash); causes failure→fix patterns to not be recognized
+- Missing shared tool files in sandbox caused SENTINEL's full chain to fail silently — always verify all tool dependencies (check-delta.js, post-comms.js, self-assess.js, update-run-state.js) are deployed before running agent cycles
+- When collect.sh fails with 'EIAP endpoint unreachable' inside sandbox, check if EIAP_URL env var is set. If empty, the script should auto-detect sandbox environment and default to http://host.openshell.internal:8090 instead of localhost.
+- Avoid: Command failed: git commit -m "$(cat <<'EOF'
+Event-driven agent cascade + direct EIAP polling — two breakthroughs... → Instead: Check command syntax, file paths, and prerequisites before running
+- Avoid: Command failed: docker exec openshell-cluster-nemoclaw kubectl exec -n openshell nka-telco -- ls -la /sandbox/.op... → Instead: Check command syntax, file paths, and prerequisites before running
+- Avoid: Command failed: cat /tmp/claude-1000/-home-ebiarao-repos-wsl-personal-projects-openclaw-autonomous-telco-agents/5... → Instead: Check command syntax, file paths, and prerequisites before running
+- Avoid: Command failed: openshell sandbox list 2>/dev/null; curl -s http://localhost:3000/api/service-health 2>&1 → Instead: Check command syntax, file paths, and prerequisites before running
+- When agent tasks complete but downstream verifications report missing output files, check if required shared tools were uploaded during sandbox initialization — re-push missing files with 'docker exec ... kubectl cp' rather than restarting the sandbox.
+- Exit code 7 from Bash commands in this project context often indicates timing issues (services not fully up) or missing sandboxes during rebuild — retry after confirming 'openshell sandbox list' shows 'Ready' state.
+- Background command 'Verify services and cron message' failed with exit code 7
+- Background command 'Verify services and cron message' failed with exit code 7
+- Avoid: Command failed: openshell status 2>/dev/null | head -5; openshell sandbox list 2>/dev/null → Instead: Check command syntax, file paths, and prerequisites before running
+- SENTINEL verify task failed - collect.sh not running after rebuild (exit code 1)
+- When tool outputs succeed but dependent processes don't trigger, check for stale state like file offsets — trigger-watcher set initial offset from old pre-rebuild state, causing it to miss new entries after rebuild+resync
+- Avoid: Command failed: curl -s http://localhost:3000/api/service-health 2>&1; echo ""; docker exec openshell-cluster-nem... → Instead: Check command syntax, file paths, and prerequisites before running
+- Avoid: Command failed: bash scripts/upgrade-sandbox.sh 2>&1 → Instead: Check command syntax, file paths, and prerequisites before running
+- Avoid: Command failed: echo "Waiting 2 min for next SENTINEL cron tick to fire with delivery:silent..." && sleep 120 && ... → Instead: Check command syntax, file paths, and prerequisites before running
+- Avoid: Command failed: kill $(pgrep -f "trigger-watcher.js") 2>/dev/null; sleep 1; rm -f artifacts/.trigger-watcher-offs... → Instead: Check command syntax, file paths, and prerequisites before running
+- Fix pairing in tool-outcome-watcher fails when the original failing command and the fix command have low string similarity (low Jaccard similarity). Pattern matching is insufficient; needs LLM reasoning about command intent.
+- Avoid: Command failed: kill $(pgrep -f "node.*webui/server.js") 2>/dev/null; sleep 1; cd /home/ebiarao/repos-wsl/persona... → Instead: Check command syntax, file paths, and prerequisites before running
+- Avoid: Command failed: ls artifacts/urgent-trigger-ORACLE.json 2>&1 → Instead: Check command syntax, file paths, and prerequisites before running
+- Avoid: Command failed: node agents/shared/tools/post-comms.js '{"agent":"TEST","type":"handoff","priority":"bad","messag... → Instead: Check command syntax, file paths, and prerequisites before running
 - Avoid: Command failed: kill $(pgrep -f "node.*webui/server.js") 2>/dev/null; kill $(pgrep -f "node.*mock-eiap/server.js"... → Instead: Check command syntax, file paths, and prerequisites before running
 - Avoid: Command failed: pgrep -af "webui/server" ; curl -s http://localhost:3000/api/service-health 2>&1 → Instead: Check command syntax, file paths, and prerequisites before running
 - Avoid: Command failed: kill $(pgrep -f "node.*webui/server.js") 2>/dev/null; sleep 1; node webui/server.js > /tmp/webui.... → Instead: Check command syntax, file paths, and prerequisites before running
-- Avoid: Command failed: claude-recall outcomes 2>&1 → Instead: Check command syntax, file paths, and prerequisites before running
 - Avoid: Command failed: npm whoami 2>&1 && npm config get registry 2>&1 → Instead: Check command syntax, file paths, and prerequisites before running
 - Claude-recall MCP Server failed to start with npx claude-recall@latest mcp start
 - Claude-recall MCP Server failed to start with command: npx -y claude-recall@latest mcp start
@@ -25,6 +50,38 @@ Auto-generated from 27 memories. Last updated: 2026-04-06.
 - Avoid: Command failed: npx jest tests/unit/failure-detectors.test.ts 2>&1 → Instead: Check command syntax, file paths, and prerequisites before running
 - SQLite query error: LIKE clause needs proper string literal syntax with single quotes in better-sqlite3
 - Node.js syntax error: multiline strings in -e flag not properly escaped; newlines break the command parsing
+- {"what_failed":"Bash command failed: git commit -m \"$(cat <<'EOF'\nEvent-driven agent cascade + direct EIAP polling — two breakthroughs...","why_failed":"Exit code 128\nerror: gpg failed to sign the data:\n[GNUPG:] KEY_CONSIDERED F3CE7255442A2E7886375262B810E846921AD84F 2\n[GNUPG:] BEGIN_SIGNING H10\n[GNUPG:] USERID_HINT B810E846921AD84F Raoul Biagioni ...","what_should_do":"Check inputs and prerequisites before retrying","context":"Bash error: Exit code 128\nerror: gpg failed to sign the data:\n[GNUPG:] KEY_CONSIDERED F3CE7255442A2E7886375262B810E846921AD84F 2\n[GNUPG:] BEGIN_SIGNING H10\n[GNUPG:] USERID_HINT B810E846921AD84F Raoul Biagioni ...","preventative_checks":["Verify tool inputs are correct","Check preconditions"]}
+- {"what_failed":"Bash command failed: docker exec openshell-cluster-nemoclaw kubectl exec -n openshell nka-telco -- ls -la /sandbox/.op...","why_failed":"Exit code 1\nfile missing","what_should_do":"Check inputs and prerequisites before retrying","context":"Bash error: Exit code 1\nfile missing","preventative_checks":["Verify tool inputs are correct","Check preconditions"]}
+- {"what_failed":"Bash command failed: cat /tmp/claude-1000/-home-ebiarao-repos-wsl-personal-projects-openclaw-autonomous-telco-agents/5...","why_failed":"Exit code 7\n\u001b[1mNAME     \u001b[0m  \u001b[1mNAMESPACE\u001b[0m  \u001b[1mCREATED            \u001b[0m  \u001b[1mPHASE\u001b[0m\nnka-telco  openshell  2026-04-06 21:16:26  \u001b[32mReady\u001b[39m\n\u001b[1mNAME     \u001b[0m  \u001b[1mNAMESPACE\u001b[0m  \u001b[1mCRE...","what_should_do":"Check inputs and prerequisites before retrying","context":"Bash error: Exit code 7\n\u001b[1mNAME     \u001b[0m  \u001b[1mNAMESPACE\u001b[0m  \u001b[1mCREATED            \u001b[0m  \u001b[1mPHASE\u001b[0m\nnka-telco  openshell  2026-04-06 21:16:26  \u001b[32mReady\u001b[39m\n\u001b[1mNAME     \u001b[0m  \u001b[1mNAMESPACE\u001b[0m  \u001b[1mCRE...","preventative_checks":["Verify tool inputs are correct","Check preconditions"]}
+- {"what_failed":"Bash command failed: openshell sandbox list 2>/dev/null; curl -s http://localhost:3000/api/service-health 2>&1","why_failed":"Exit code 7\nNo sandboxes found.","what_should_do":"Check inputs and prerequisites before retrying","context":"Bash error: Exit code 7\nNo sandboxes found.","preventative_checks":["Verify tool inputs are correct","Check preconditions"]}
+- {"what_failed":"Bash command failed: docker exec openshell-cluster-nemoclaw kubectl exec -n openshell nka-telco -- bash -c 'echo $EIAP...","why_failed":"Exit code 1\n\n---\n[collect] fetching EIAP endpoints at 2026-04-06T20:58:54Z\ncommand terminated with exit code 1","what_should_do":"Check inputs and prerequisites before retrying","context":"Bash error: Exit code 1\n\n---\n[collect] fetching EIAP endpoints at 2026-04-06T20:58:54Z\ncommand terminated with exit code 1","preventative_checks":["Verify tool inputs are correct","Check preconditions"]}
+- {"what_failed":"Bash command failed: openshell status 2>/dev/null | head -5; openshell sandbox list 2>/dev/null","why_failed":"Exit code 1\n\u001b[1m\u001b[36mGateway Status\u001b[39m\u001b[0m\n\n  \u001b[2mStatus:\u001b[0m No gateway configured.\n\nDeploy a gateway with: \u001b[2mopenshell gateway start\u001b[0m","what_should_do":"Check inputs and prerequisites before retrying","context":"Bash error: Exit code 1\n\u001b[1m\u001b[36mGateway Status\u001b[39m\u001b[0m\n\n  \u001b[2mStatus:\u001b[0m No gateway configured.\n\nDeploy a gateway with: \u001b[2mopenshell gateway start\u001b[0m","preventative_checks":["Verify tool inputs are correct","Check preconditions"]}
+- {"what_failed":"Read failed","why_failed":"File content (12314 tokens) exceeds maximum allowed tokens (10000). Use offset and limit parameters to read specific portions of the file, or search for specific content instead of reading the whol...","what_should_do":"Check inputs and prerequisites before retrying","context":"Error: File content (12314 tokens) exceeds maximum allowed tokens (10000). Use offset and limit parameters to read specific portions of the file, or search for specific content instead of reading the whol...","preventative_checks":["Verify tool inputs are correct","Check preconditions"]}
+- {"what_failed":"Bash command failed: B64=$(base64 -w0 < agents/shared/tools/post-comms.js) && ssh -q -o StrictHostKeyChecking=no -o Us...","why_failed":"Exit code 1","what_should_do":"Check inputs and prerequisites before retrying","context":"Bash error: Exit code 1","preventative_checks":["Verify tool inputs are correct","Check preconditions"]}
+- {"what_failed":"Read failed","why_failed":"File does not exist. Note: your current working directory is /home/ebiarao/repos-wsl/personal-projects/openclaw-autonomous-telco-agents.","what_should_do":"Check inputs and prerequisites before retrying","context":"Error: File does not exist. Note: your current working directory is /home/ebiarao/repos-wsl/personal-projects/openclaw-autonomous-telco-agents.","preventative_checks":["Verify tool inputs are correct","Check preconditions"]}
+- {"what_failed":"Bash command failed: docker exec openshell-cluster-nemoclaw kubectl exec -n openshell nka-telco -- tail -10 /sandbox/....","why_failed":"Exit code 1\nTraceback (most recent call last):\n  File \"<string>\", line 6, in <module>\n    d=json.loads(line)\n  File \"/home/ebiarao/miniconda3/lib/python3.13/json/__init__.py\", line 346, in loads\n  ...","what_should_do":"Check inputs and prerequisites before retrying","context":"Bash error: Exit code 1\nTraceback (most recent call last):\n  File \"<string>\", line 6, in <module>\n    d=json.loads(line)\n  File \"/home/ebiarao/miniconda3/lib/python3.13/json/__init__.py\", line 346, in loads\n  ...","preventative_checks":["Verify tool inputs are correct","Check preconditions"]}
+- {"what_failed":"Bash command failed: pgrep -af trigger-watcher; echo \"---\"; cat artifacts/trigger-log.jsonl 2>/dev/null || echo \"no tr...","why_failed":"Exit code 1\n3180787 node /home/ebiarao/repos-wsl/personal-projects/openclaw-autonomous-telco-agents/scripts/trigger-watcher.js\n3233721 /bin/bash -c source /home/ebiarao/.claude/shell-snapshots/snap...","what_should_do":"Check inputs and prerequisites before retrying","context":"Bash error: Exit code 1\n3180787 node /home/ebiarao/repos-wsl/personal-projects/openclaw-autonomous-telco-agents/scripts/trigger-watcher.js\n3233721 /bin/bash -c source /home/ebiarao/.claude/shell-snapshots/snap...","preventative_checks":["Verify tool inputs are correct","Check preconditions"]}
+- {"what_failed":"Bash command failed: cat artifacts/trigger-log.jsonl 2>/dev/null || echo \"(no triggers yet)\"; echo \"===\"; tail -5 arti...","why_failed":"Exit code 1\n(no triggers yet)\n===","what_should_do":"Check inputs and prerequisites before retrying","context":"Bash error: Exit code 1\n(no triggers yet)\n===","preventative_checks":["Verify tool inputs are correct","Check preconditions"]}
+- {"what_failed":"Bash command failed: pwd; ls agents/nka/scripts/collect.sh 2>/dev/null","why_failed":"Exit code 2\n/home/ebiarao/repos-wsl/personal-projects/openclaw-autonomous-telco-agents/webui/client","what_should_do":"Check inputs and prerequisites before retrying","context":"Bash error: Exit code 2\n/home/ebiarao/repos-wsl/personal-projects/openclaw-autonomous-telco-agents/webui/client","preventative_checks":["Verify tool inputs are correct","Check preconditions"]}
+- {"what_failed":"Bash command failed: ls artifacts/ 2>/dev/null; mkdir -p artifacts && EIAP_URL=http://127.0.0.1:8090 bash agents/nka/s...","why_failed":"Exit code 127\nbash: agents/nka/scripts/collect.sh: No such file or directory","what_should_do":"Check inputs and prerequisites before retrying","context":"Bash error: Exit code 127\nbash: agents/nka/scripts/collect.sh: No such file or directory","preventative_checks":["Verify tool inputs are correct","Check preconditions"]}
+- {"what_failed":"Bash command failed: cat artifacts/state.json","why_failed":"Exit code 1\ncat: artifacts/state.json: No such file or directory","what_should_do":"Check inputs and prerequisites before retrying","context":"Bash error: Exit code 1\ncat: artifacts/state.json: No such file or directory","preventative_checks":["Verify tool inputs are correct","Check preconditions"]}
+- {"what_failed":"Bash command failed: sleep 3 && curl -s http://localhost:3000/api/service-health","why_failed":"Exit code 7","what_should_do":"Check inputs and prerequisites before retrying","context":"Bash error: Exit code 7","preventative_checks":["Verify tool inputs are correct","Check preconditions"]}
+- {"what_failed":"Bash command failed: cd /home/ebiarao/repos-wsl/personal-projects/openclaw-autonomous-telco-agents && kill $(pgrep -f ...","why_failed":"Exit code 144","what_should_do":"Check inputs and prerequisites before retrying","context":"Bash error: Exit code 144","preventative_checks":["Verify tool inputs are correct","Check preconditions"]}
+- {"what_failed":"Bash command failed: node -e \"\nconst fs = require('fs');\nconst wsFile = 'mock-eiap/world/world-state.json';\nconst ws =...","why_failed":"Exit code 1\nnode:fs:449\n    return binding.readFileUtf8(path, stringToFlags(options.flag));\n                   ^\n\nError: ENOENT: no such file or directory, open 'mock-eiap/world/world-state.json'\n ...","what_should_do":"Check inputs and prerequisites before retrying","context":"Bash error: Exit code 1\nnode:fs:449\n    return binding.readFileUtf8(path, stringToFlags(options.flag));\n                   ^\n\nError: ENOENT: no such file or directory, open 'mock-eiap/world/world-state.json'\n ...","preventative_checks":["Verify tool inputs are correct","Check preconditions"]}
+- {"what_failed":"Bash command failed: curl -s http://localhost:3000/api/service-health 2>&1; echo \"\"; docker exec openshell-cluster-nem...","why_failed":"Exit code 1\n{\"eiap\":true,\"dataStale\":false,\"signalsAge\":3,\"checkedAt\":\"2026-04-06T18:54:08.040Z\"}\nTraceback (most recent call last):\n  File \"<string>\", line 6, in <module>\n    print(f\"{j['name']:12...","what_should_do":"Check inputs and prerequisites before retrying","context":"Bash error: Exit code 1\n{\"eiap\":true,\"dataStale\":false,\"signalsAge\":3,\"checkedAt\":\"2026-04-06T18:54:08.040Z\"}\nTraceback (most recent call last):\n  File \"<string>\", line 6, in <module>\n    print(f\"{j['name']:12...","preventative_checks":["Verify tool inputs are correct","Check preconditions"]}
+- {"what_failed":"Bash command failed: docker exec openshell-cluster-nemoclaw kubectl exec -n openshell nka-telco -- bash -c 'openclaw c...","why_failed":"Exit code 1\nError: gateway closed (1006 abnormal closure (no close frame)): no close reason\nGateway target: ws://127.0.0.1:18789\nSource: local loopback\nConfig: /root/.openclaw/openclaw.json\nBind: l...","what_should_do":"Check inputs and prerequisites before retrying","context":"Bash error: Exit code 1\nError: gateway closed (1006 abnormal closure (no close frame)): no close reason\nGateway target: ws://127.0.0.1:18789\nSource: local loopback\nConfig: /root/.openclaw/openclaw.json\nBind: l...","preventative_checks":["Verify tool inputs are correct","Check preconditions"]}
+- {"what_failed":"Bash command failed: docker exec openshell-cluster-nemoclaw kubectl exec -n openshell nka-telco -- bash -c 'ps aux | g...","why_failed":"Exit code 1\nbash: line 1: ps: command not found\ncommand terminated with exit code 1","what_should_do":"Check inputs and prerequisites before retrying","context":"Bash error: Exit code 1\nbash: line 1: ps: command not found\ncommand terminated with exit code 1","preventative_checks":["Verify tool inputs are correct","Check preconditions"]}
+- {"what_failed":"Bash command failed: docker exec openshell-cluster-nemoclaw kubectl exec -n openshell nka-telco -- bash -c 'openclaw c...","why_failed":"Exit code 1\nError: gateway closed (1006 abnormal closure (no close frame)): no close reason\nGateway target: ws://127.0.0.1:18789\nSource: local loopback\nConfig: /root/.openclaw/openclaw.json\nBind: l...","what_should_do":"Check inputs and prerequisites before retrying","context":"Bash error: Exit code 1\nError: gateway closed (1006 abnormal closure (no close frame)): no close reason\nGateway target: ws://127.0.0.1:18789\nSource: local loopback\nConfig: /root/.openclaw/openclaw.json\nBind: l...","preventative_checks":["Verify tool inputs are correct","Check preconditions"]}
+- {"what_failed":"Bash command failed: docker exec openshell-cluster-nemoclaw kubectl exec -n openshell nka-telco -- bash -c 'openclaw c...","why_failed":"Exit code 1\ncommand terminated with exit code 1","what_should_do":"Check inputs and prerequisites before retrying","context":"Bash error: Exit code 1\ncommand terminated with exit code 1","preventative_checks":["Verify tool inputs are correct","Check preconditions"]}
+- Fix pairing based on Jaccard string similarity between failing command and fix command is too weak when fixes look syntactically different from original commands (e.g. adding echo pipe vs original bash call); consider semantic matching or explicit fix linking
+- {"what_failed":"Bash command failed: bash scripts/upgrade-sandbox.sh 2>&1","why_failed":"Exit code 1\n\n\u001b[0;32m[upgrade]\u001b[0m ==========================================\n\u001b[0;32m[upgrade]\u001b[0m   NeMoClaw Sandbox Rebuild\n\u001b[0;32m[upgrade]\u001b[0m ==========================================\n\n\u001b[0;32m...","what_should_do":"Check inputs and prerequisites before retrying","context":"Bash error: Exit code 1\n\n\u001b[0;32m[upgrade]\u001b[0m ==========================================\n\u001b[0;32m[upgrade]\u001b[0m   NeMoClaw Sandbox Rebuild\n\u001b[0;32m[upgrade]\u001b[0m ==========================================\n\n\u001b[0;32m...","preventative_checks":["Verify tool inputs are correct","Check preconditions"]}
+- {"what_failed":"Bash command failed: echo \"Waiting 2 min for next SENTINEL cron tick to fire with delivery:silent...\" && sleep 120 && ...","why_failed":"Exit code 1\nWaiting 2 min for next SENTINEL cron tick to fire with delivery:silent...\nTraceback (most recent call last):\n  File \"<string>\", line 1, in <module>\n    import json,sys;d=json.load(sys.s...","what_should_do":"Check inputs and prerequisites before retrying","context":"Bash error: Exit code 1\nWaiting 2 min for next SENTINEL cron tick to fire with delivery:silent...\nTraceback (most recent call last):\n  File \"<string>\", line 1, in <module>\n    import json,sys;d=json.load(sys.s...","preventative_checks":["Verify tool inputs are correct","Check preconditions"]}
+- {"what_failed":"Bash command failed: pkill -f \"node.*trigger-watcher\" 2>/dev/null; sleep 1; rm -f /tmp/trigger-watcher.log artifacts/....","why_failed":"Exit code 144","what_should_do":"Check inputs and prerequisites before retrying","context":"Bash error: Exit code 144","preventative_checks":["Verify tool inputs are correct","Check preconditions"]}
+- {"what_failed":"Bash command failed: pkill -f \"trigger-watcher\" 2>/dev/null; sleep 1; rm -f /tmp/trigger-watcher.log artifacts/.trigge...","why_failed":"Exit code 144","what_should_do":"Check inputs and prerequisites before retrying","context":"Bash error: Exit code 144","preventative_checks":["Verify tool inputs are correct","Check preconditions"]}
+- {"what_failed":"Bash command failed: kill $(pgrep -f \"trigger-watcher.js\") 2>/dev/null; sleep 1; rm -f artifacts/.trigger-watcher-offs...","why_failed":"Exit code 144","what_should_do":"Check inputs and prerequisites before retrying","context":"Bash error: Exit code 144","preventative_checks":["Verify tool inputs are correct","Check preconditions"]}
+- {"what_failed":"Bash command failed: openshell sandbox list 2>/dev/null; openshell status 2>/dev/null; ls ~/.openclaw-ata/cron/jobs.js...","why_failed":"Exit code 2\n\u001b[1mNAME     \u001b[0m  \u001b[1mNAMESPACE\u001b[0m  \u001b[1mCREATED            \u001b[0m  \u001b[1mPHASE\u001b[0m\nnka-telco  openshell  2026-04-06 11:22:57  \u001b[32mReady\u001b[39m\n\u001b[1m\u001b[36mServer Status\u001b[39m\u001b[0m\n\n  \u001b[2mGatewa...","what_should_do":"Check inputs and prerequisites before retrying","context":"Bash error: Exit code 2\n\u001b[1mNAME     \u001b[0m  \u001b[1mNAMESPACE\u001b[0m  \u001b[1mCREATED            \u001b[0m  \u001b[1mPHASE\u001b[0m\nnka-telco  openshell  2026-04-06 11:22:57  \u001b[32mReady\u001b[39m\n\u001b[1m\u001b[36mServer Status\u001b[39m\u001b[0m\n\n  \u001b[2mGatewa...","preventative_checks":["Verify tool inputs are correct","Check preconditions"]}
+- {"what_failed":"Bash command failed: ls artifacts/urgent-trigger-ORACLE.json 2>&1","why_failed":"Exit code 2\nls: cannot access 'artifacts/urgent-trigger-ORACLE.json': No such file or directory","what_should_do":"Check inputs and prerequisites before retrying","context":"Bash error: Exit code 2\nls: cannot access 'artifacts/urgent-trigger-ORACLE.json': No such file or directory","preventative_checks":["Verify tool inputs are correct","Check preconditions"]}
+- {"what_failed":"Bash command failed: echo \"38\" > artifacts/.trigger-watcher-offset && timeout 8 node scripts/trigger-watcher.js 2>&1","why_failed":"Exit code 124\n[trigger-watcher] 2026-04-06T18:17:44.804Z started — polling every 5s\n[trigger-watcher] 2026-04-06T18:17:44.806Z initial offset: 39 lines","what_should_do":"Check inputs and prerequisites before retrying","context":"Bash error: Exit code 124\n[trigger-watcher] 2026-04-06T18:17:44.804Z started — polling every 5s\n[trigger-watcher] 2026-04-06T18:17:44.806Z initial offset: 39 lines","preventative_checks":["Verify tool inputs are correct","Check preconditions"]}
 - {"what_failed":"Bash command failed: node webui/server.js > /tmp/webui.log 2>&1 & sleep 3 && curl -s http://localhost:3000/api/service...","why_failed":"Exit code 7\n[1]+  Exit 1                  node webui/server.js > /tmp/webui.log 2>&1","what_should_do":"Check inputs and prerequisites before retrying","context":"Bash error: Exit code 7\n[1]+  Exit 1                  node webui/server.js > /tmp/webui.log 2>&1","preventative_checks":["Verify tool inputs are correct","Check preconditions"]}
 - {"what_failed":"Bash command failed: sleep 2 && curl -s http://localhost:3000/api/service-health 2>&1","why_failed":"Exit code 7","what_should_do":"Check inputs and prerequisites before retrying","context":"Bash error: Exit code 7","preventative_checks":["Verify tool inputs are correct","Check preconditions"]}
 - {"what_failed":"Bash command failed: kill $(pgrep -f \"node.*webui/server.js\") 2>/dev/null; sleep 1; cd /home/ebiarao/repos-wsl/persona...","why_failed":"Exit code 144","what_should_do":"Check inputs and prerequisites before retrying","context":"Bash error: Exit code 144","preventative_checks":["Verify tool inputs are correct","Check preconditions"]}
@@ -34,10 +91,14 @@ Auto-generated from 27 memories. Last updated: 2026-04-06.
 - {"what_failed":"Read failed","why_failed":"File content (11505 tokens) exceeds maximum allowed tokens (10000). Use offset and limit parameters to read specific portions of the file, or search for specific content instead of reading the whol...","what_should_do":"Check inputs and prerequisites before retrying","context":"Error: File content (11505 tokens) exceeds maximum allowed tokens (10000). Use offset and limit parameters to read specific portions of the file, or search for specific content instead of reading the whol...","preventative_checks":["Verify tool inputs are correct","Check preconditions"]}
 - Avoid: Test command reported failures: npx jest tests/unit/failure-detectors.test.ts 2>&1 → Instead: Read test output carefully — exit code 0 does not mean all tests passed
 - Node.js -e flag cannot parse multiline strings with unescaped newlines in single quotes
+- Grep with escaped quotes (pattern="SESSION_EXTRACTION_PROMPT\"") returns no matches — use unescaped pattern="SESSION_EXTRACTION_PROMPT" instead
+- Gateway token mismatch error — requires verifying token in Control UI settings via dashboard URL
 - claude-recall reconnection failed after reinstall
 - claude-recall reconnection failed after reinstall attempt
 - Output is showing '>' instead of expected response
 - Code output returns '>' instead of expected response
+- Background command 'Wait one more SENTINEL cycle' failed with exit code 1
+- Avoid: Claude backtracked: "Let me try another" → Instead: Consider the problem more carefully before starting. Check for similar past failures.
 - Avoid: Claude backtracked: "let me try a different approach" → Instead: Consider the problem more carefully before starting. Check for similar past failures.
 
 ---

package/.claude/skills/auto-failure-lessons/manifest.json

@@ -1,15 +1,39 @@
 {
   "topicId": "failure-lessons",
-  "sourceHash": "543a967cdc5416d35aa4d95b097e319a3d4c184780ea4ee2975551d2c1187e59",
-  "memoryCount": 27,
-  "generatedAt": "2026-04-06T18:15:28.298Z",
+  "sourceHash": "0a433bc57da8ed5c3137f8676cae1ec253e4a5037b115400d0e0c9e394789440",
+  "memoryCount": 87,
+  "generatedAt": "2026-04-06T21:44:03.038Z",
   "memoryKeys": [
     "promoted_1775498319002_awwrsovw6",
+    "promoted_1775499579486_bxvgjc80r",
+    "hook_failure_1775502039213_k4yusuan1",
     "hook_failure_1772637584921_0tj4rrxnt",
+    "hook_failure_1775503464385_vz5vdzxvs",
+    "hook_failure_1775500959351_j74owpvko",
+    "hook_failure_1775511109813_ly4cjtde6",
+    "hook_failure_1775509498191_f79imbrs4",
+    "hook_failure_non-zero-exit_1775511575363_nen5eyl3b",
+    "hook_failure_non-zero-exit_1775510785418_2xm52g5q8",
+    "hook_failure_non-zero-exit_1775510309712_olv4egt1g",
+    "hook_failure_non-zero-exit_1775510309694_1cipsasft",
+    "hook_failure_1775510785331_kxr1uoozx",
+    "hook_failure_1775510309605_z58f30w5t",
+    "hook_failure_1775510305548_qq8hdyk12",
+    "hook_failure_1775510298887_aoeh4m2cl",
+    "hook_failure_non-zero-exit_1775508193513_ms20rhkf1",
+    "hook_failure_1775509104206_lg9fvtipb",
+    "hook_failure_1775504451387_fmycaea7m",
+    "hook_failure_non-zero-exit_1775503134951_mq2x1hj77",
+    "hook_failure_non-zero-exit_1775502651250_4ah16usm4",
+    "hook_failure_non-zero-exit_1775502651238_l0c5dkykp",
+    "hook_failure_non-zero-exit_1775501919378_s3j17tkr3",
+    "hook_failure_1775500665745_zl6zsp7jw",
+    "hook_failure_non-zero-exit_1775499579462_n5xb6rr8r",
+    "hook_failure_non-zero-exit_1775499579440_2tv110hs9",
+    "hook_failure_non-zero-exit_1775499579425_n53js176o",
     "hook_failure_non-zero-exit_1775498318987_pqbmsatub",
     "hook_failure_non-zero-exit_1775498318976_mzbyvdceo",
     "hook_failure_non-zero-exit_1775498318963_thiuiiqef",
-    "hook_failure_non-zero-exit_1774020949485_6ubuoswae",
     "hook_failure_non-zero-exit_1773409269877_ful451241",
     "hook_failure_1773409616313_gmsfcbuzh",
     "hook_failure_1773409607268_83ie5yunz",
@@ -17,6 +41,38 @@
     "hook_failure_non-zero-exit_1772640279977_g3gwlfoqi",
     "hook_failure_1772637570984_yxs8zmurp",
     "hook_failure_1772637485532_djep8eysa",
+    "hook_failure_1775511332509_3mxf5vtg3",
+    "hook_failure_1775510710817_sn23160zp",
+    "hook_failure_1775510222535_o2t7vm7qp",
+    "hook_failure_1775510141049_oo7v7htpv",
+    "hook_failure_1775509134627_wu9wz6w1p",
+    "hook_failure_1775508183011_gy3nzxwiy",
+    "hook_failure_1775505258754_l50o6be33",
+    "hook_failure_1775503921254_zyjka5vj9",
+    "hook_failure_1775503659141_1fh62brao",
+    "hook_failure_1775503198240_qzmm07ves",
+    "hook_failure_1775503190006_k5mdxukzl",
+    "hook_failure_1775503183507_rs20xzomb",
+    "hook_failure_1775503035650_efta0k28k",
+    "hook_failure_1775503031427_e9070twhb",
+    "hook_failure_1775503009947_ea2euap40",
+    "hook_failure_1775502961298_snolsqnu9",
+    "hook_failure_1775502951488_hyvabjgp1",
+    "hook_failure_1775502809370_icsa9nxsy",
+    "hook_failure_1775501648406_78lsvrrb1",
+    "hook_failure_1775501136927_v2zesbrwh",
+    "hook_failure_1775501122267_ejvqdr0id",
+    "hook_failure_1775501085124_c92qa38u7",
+    "hook_failure_1775501078345_pryhog20w",
+    "hook_failure_1775500618695_6grbwpm3t",
+    "hook_failure_1775500273633_yrw798ahh",
+    "hook_failure_1775500125145_ad65bw8rs",
+    "hook_failure_1775499854068_ze23fp9gf",
+    "hook_failure_1775499797961_6ziqluf1i",
+    "hook_failure_1775499784048_qsshv04hl",
+    "hook_failure_1775499638896_pfhgi3kn5",
+    "hook_failure_1775499503779_k9vabnkif",
+    "hook_failure_1775499472886_1exmgb4ef",
     "hook_failure_1775499266183_drme5d0j6",
     "hook_failure_1775499258759_4unleb4yv",
     "hook_failure_1775499252924_czqqpfsp3",
@@ -26,10 +82,14 @@
     "hook_failure_1775498542480_wc23ux8gt",
     "hook_failure_silent-test-failure_1772640279996_m8f3ks8fw",
     "hook_failure_1772637495154_i21dho3dv",
+    "hook_failure_1775511713477_psti3hr7y",
+    "hook_failure_1775502085189_pnbd0kkg1",
     "hook_failure_1773410663873_oegccxk83",
     "hook_failure_1773410656672_kno1aw8wf",
     "hook_failure_1772637391602_cjj3esg33",
     "hook_failure_1772637376157_hvco6jbm4",
+    "hook_failure_1775509857057_847nm61ld",
+    "hook_failure_backtracking_1775508193524_54wn8o688",
     "hook_failure_backtracking_1772640280006_snb0x5cgg"
   ]
 }

package/.claude/skills/auto-preferences/SKILL.md

@@ -8,91 +8,33 @@ source: claude-recall
 
 # Preferences
 
-Auto-generated from 81 memories. Last updated: 2026-04-06.
+Auto-generated from 23 memories. Last updated: 2026-04-06.
 
 ## Rules
 
+- Session test preference 1775511843229
+- Test preference 1775511843109-2
+- Test preference 1775511843109-1
+- Test preference 1775511843109-0
+- Test memory content
+- Session test preference 1775500740595
+- Test preference 1775500740495-2
+- Test preference 1775500740495-1
+- Test preference 1775500740495-0
+- Test memory content
 - Session test preference 1775499328559
 - Test preference 1775499328384-2
 - Test preference 1775499328384-1
 - Test preference 1775499328384-0
 - Test memory content
 - Session test preference 1775496241643
-- Test preference 1775496241514-2
-- Test preference 1775496241514-1
-- Test preference 1775496241514-0
-- Test memory content
-- Session test preference 1775494279149
-- Test preference 1775494279061-2
-- Test preference 1775494279061-1
-- Test preference 1775494279061-0
-- Test memory content
-- Session test preference 1775492069465
-- Test preference 1775492069353-2
-- Test preference 1775492069353-1
-- Test preference 1775492069353-0
-- Test memory content
-- Session test preference 1775491767519
-- Test preference 1775491767395-2
-- Test preference 1775491767395-1
-- Test preference 1775491767395-0
-- Test memory content
 - After each refactoring, document the changes made. Don't batch documentation to the end — write it as you go.
 - Any major refactoring requires exhaustive search to make sure nothing is missed. Always grep/search comprehensively before and after changes to verify no stale references, broken imports, or missed files remain.
-- Session test preference 1775169786712
-- Test preference 1775169786565-2
-- Test preference 1775169786565-1
-- Test preference 1775169786565-0
-- Test memory content
-- Session test preference 1775169704849
-- Test preference 1775169704668-2
-- Test preference 1775169704668-1
-- Test preference 1775169704668-0
-- Test memory content
-- Session test preference 1775169639212
-- Test preference 1775169639121-2
-- Test preference 1775169639121-1
-- Test preference 1775169639121-0
-- Test memory content
-- Session test preference 1775169273366
-- Test preference 1775169273296-2
-- Test preference 1775169273296-1
-- Test preference 1775169273296-0
-- Test memory content
-- Session test preference 1775169256189
-- Test preference 1775169256133-2
-- Test preference 1775169256133-1
-- Test preference 1775169256133-0
-- Test memory content
-- Session test preference 1775169244752
-- Test preference 1775169244686-2
-- Test preference 1775169244686-1
-- Test preference 1775169244686-0
-- Test memory content
 - Markdown documentation files should be named with the convention YYYYMMDD_<filename>.md (e.g., 20260402_nemoclaw_integration.md). Use underscores, not hyphens, in the date-prefixed filenames.
-- Session test preference 1775160781884
-- Test preference 1775160781820-2
-- Test preference 1775160781820-1
-- Test preference 1775160781820-0
-- Test memory content
-- Session test preference 1775159399698
-- Test preference 1775159399559-2
-- Test preference 1775159399559-1
-- Test preference 1775159399559-0
-- Test memory content
-- Session test preference 1775157082982
-- Test preference 1775157082853-2
-- Test preference 1775157082853-1
-- Test preference 1775157082853-0
-- Test memory content
-- Session test preference 1775156381211
-- Test preference 1775156381110-2
-- Test preference 1775156381110-1
-- Test preference 1775156381110-0
-- Test memory content
-- axios npm package was compromised in a supply chain attack (axios@1.14.1 pulled in malicious plain-crypto-js@4.2.1). Claude Recall is NOT affected — does not use axios. Verified 2026-04-01. If axios is ever added as a dependency, pin the version and audit lockfiles.
+- Always ask before deleting anything
+- When user questions the scope or complexity of a solution mid-session, pause and clarify the reusable abstractions extracted from the implementation. Identify generalizable patterns (e.g., trigger-watcher, signal-based coordination) that justify the engineering effort beyond the immediate use case.
+- When discovering architectural issues, create comprehensive documentation immediately (host/sandbox duality) and add redesign work to TODO list for later iteration.
 - For agentic design, agents should decide when to run rather than using cron jobs for scheduling
-- Upgrade all projects whenever a new version is pushed
 
 ---
 *Auto-generated by Claude Recall. Regenerate: `npx claude-recall skills generate`*

package/.claude/skills/auto-preferences/manifest.json

@@ -1,89 +1,31 @@
 {
   "topicId": "preferences",
-  "sourceHash": "dce1646010683da840451f75aec6672ca62e50210bc5d2c89fe733ab9f9a60bd",
-  "memoryCount": 81,
-  "generatedAt": "2026-04-06T18:15:28.596Z",
+  "sourceHash": "819c9fa966f5ffabe7b065fa6290958f1601072fb16e47d162bd513228ec08d5",
+  "memoryCount": 23,
+  "generatedAt": "2026-04-06T21:44:03.259Z",
   "memoryKeys": [
+    "memory_1775511843231_at7wjj6kf",
+    "memory_1775511843181_7ruo2iof1",
+    "memory_1775511843150_vdcha6qwp",
+    "memory_1775511843112_mmaq6zmrn",
+    "memory_1775511842984_cn67mk6m3",
+    "memory_1775500740596_xok3iey9j",
+    "memory_1775500740545_op5713e5i",
+    "memory_1775500740519_ln1fd8ow4",
+    "memory_1775500740497_f06v303gf",
+    "memory_1775500740399_ud5aghy9a",
     "memory_1775499328561_9m7ao3h8s",
     "memory_1775499328480_e2zx9g5ly",
     "memory_1775499328425_plvaaxd69",
     "memory_1775499328386_dqcav3itu",
     "memory_1775499328258_cilhe1owj",
     "memory_1775496241645_xtz7cwan4",
-    "memory_1775496241576_j7sxxuge3",
-    "memory_1775496241542_rg3hj24ud",
-    "memory_1775496241516_y3inz8rlf",
-    "memory_1775496241411_vwdz6igrm",
-    "memory_1775494279150_n4pq7zy11",
-    "memory_1775494279108_6hbe8qoit",
-    "memory_1775494279088_nv8hjdm7s",
-    "memory_1775494279062_jx4wrwn6s",
-    "memory_1775494278982_fsc491z41",
-    "memory_1775492069467_5cturlg0a",
-    "memory_1775492069400_icg4tjivf",
-    "memory_1775492069377_goix7nu9v",
-    "memory_1775492069354_wma3zh5i7",
-    "memory_1775492069258_q9d2k28wt",
-    "memory_1775491767523_p2xtn4uak",
-    "memory_1775491767447_q1dwsdfk3",
-    "memory_1775491767421_vgntf4jt8",
-    "memory_1775491767397_f181w5lqd",
-    "memory_1775491767290_s7ntmkwpg",
     "memory_1775208934902_2kovciriy",
     "memory_1775208477621_fqa3w21j1",
-    "memory_1775169786717_1zmwoe6ai",
-    "memory_1775169786630_rdudb8hbc",
-    "memory_1775169786589_zurej1v51",
-    "memory_1775169786567_3esvhc01j",
-    "memory_1775169786460_vugzw5yl0",
-    "memory_1775169704852_ezdebznpz",
-    "memory_1775169704771_ak9vkaswc",
-    "memory_1775169704739_dg78h7lte",
-    "memory_1775169704670_f1yhtrmsz",
-    "memory_1775169704515_lby33vw9c",
-    "memory_1775169639213_w3i85h357",
-    "memory_1775169639170_iewh2evi0",
-    "memory_1775169639145_x9s0519l7",
-    "memory_1775169639123_fa53agp0a",
-    "memory_1775169639037_f290kffwd",
-    "memory_1775169273368_c1e5jdl9r",
-    "memory_1775169273337_wozxe12k3",
-    "memory_1775169273319_3o1ww90na",
-    "memory_1775169273297_oqmyxziav",
-    "memory_1775169273232_eoymf3ezj",
-    "memory_1775169256190_681x8frvt",
-    "memory_1775169256164_kqderurux",
-    "memory_1775169256150_0nqbk7tx9",
-    "memory_1775169256134_ftj0m44bb",
-    "memory_1775169256069_dywd646f7",
-    "memory_1775169244754_sjfarlk41",
-    "memory_1775169244722_mavida9fg",
-    "memory_1775169244707_nmmu0s0lu",
-    "memory_1775169244688_03q4mmcky",
-    "memory_1775169244628_7kuhf5met",
     "memory_1775168901303_o8o0i4952",
-    "memory_1775160781885_fkbnc0s85",
-    "memory_1775160781857_gj5y299dr",
-    "memory_1775160781842_0xmv146n5",
-    "memory_1775160781821_o9ipt8nsc",
-    "memory_1775160781747_70e5gbp7j",
-    "memory_1775159399702_yg6ynd05n",
-    "memory_1775159399631_fmy4s2vhr",
-    "memory_1775159399606_xymkvcjfv",
-    "memory_1775159399561_o6vpb3fto",
-    "memory_1775159399463_0pvhuixyr",
-    "memory_1775157082984_oiyt9u65w",
-    "memory_1775157082927_29d7mm59z",
-    "memory_1775157082892_o9i0jbw76",
-    "memory_1775157082855_zlbl2ib0j",
-    "memory_1775157082743_fu07oklto",
-    "memory_1775156381212_ptlo01224",
-    "memory_1775156381158_3p6goornd",
-    "memory_1775156381132_ikx95mk0n",
-    "memory_1775156381111_1x4fxfu9n",
-    "memory_1775156381012_3tw7we7cz",
-    "memory_1775154208266_arwo2fctx",
-    "hook_preference_1775498374200_6d21qj14y",
-    "hook_preference_1774106575282_45sk4ep52"
+    "hook_preference_1775511561137_rjm92h4o3",
+    "hook_preference_1775508193449_aphyh28z9",
+    "hook_preference_1775503672864_weqgjstda",
+    "hook_preference_1775498374200_6d21qj14y"
   ]
 }

package/README.md

@@ -84,11 +84,11 @@ Once installed, Claude Recall works automatically in the background:
 1. **Session start** — active rules are loaded before the first action. In Claude Code, this happens via the `search_enforcer` hook; in Pi, rules are injected into the system prompt automatically
 2. **As you work** — every prompt is classified for corrections and preferences. Natural statements like *"we use tabs here"* or *"no, put tests in `__tests__/`"* are detected and stored
 3. **Tool outcomes** — results from all tools (Bash, Edit, Write, and more) are captured. Failures are stored as memories; Bash failures are paired with successful fixes
-4. **End of session** — session episodes are created, candidate lessons extracted from failures, and a promotion cycle graduates validated patterns into active rules. A session extraction pass sends the last 50 transcript entries to Haiku to extract durable project knowledge from long coding sessions
+4. **End of session** — session episodes are created, candidate lessons extracted from failures, and a promotion cycle graduates validated patterns into active rules. A session extraction pass sends the last 50 transcript entries to Haiku to identify cause-and-effect patterns: what failed, why, and what fixed it
 5. **Reask detection** — frustration signals ("still broken", "that didn't work") are recorded as outcome events
 6. **Before context compression** — aggressive memory sweep captures important context before the window shrinks
 7. **After context compression** (Claude Code only) — rules are automatically re-injected into context so they're not lost when the window shrinks
-8. **Multi-agent outcomes** — subagent completions (completed/failed/killed) are captured from task notifications and recorded as outcome events
+8. **Sub-agent recall** (Claude Code only) — when sub-agents are spawned, active rules are injected into their context automatically. Sub-agent outcomes (completed/failed/killed) are captured as events
 9. **Rules sync** (Claude Code only) — top 30 rules are exported as typed `.md` files to Claude Code's native memory directory
 
 Classification uses Claude Haiku (via `ANTHROPIC_API_KEY`) with silent regex fallback. No configuration needed.

package/dist/hooks/correction-detector.js

@@ -48,8 +48,8 @@ async function handleCorrectionDetector(input) {
     // Reject short/garbage extracts and raw dumps (not clean rules)
     if (result.extract.length < 10 || result.extract.length > 200)
         return;
-    // Corrections and preferences need high confidence; others need moderate
-    if ((result.type === 'correction' || result.type === 'preference') && result.confidence < 0.7)
+    // Corrections, preferences, and devops need high confidence to prevent noise
+    if ((result.type === 'correction' || result.type === 'preference' || result.type === 'devops') && result.confidence < 0.75)
         return;
     if (result.confidence < 0.6)
         return;

package/dist/hooks/llm-classifier.js

@@ -18,7 +18,7 @@ const SYSTEM_PROMPT = `You are a memory classifier for a developer tool. Classif
 - correction: User correcting a mistake ("no, use X not Y", "wrong, it should be...")
 - preference: User stating a clear, reusable directive about how they want things done going forward ("we use tabs", "always use TypeScript", "I prefer X"). Must be a rule that applies beyond this conversation. NOT observations, complaints, questions, debugging statements, or one-off instructions like "fix this" or "tell me about X"
 - failure: Something broke or failed ("build failed", "error in deployment")
-- devops: CI/CD, deployment, Docker, git workflow patterns
+- devops: CI/CD, deployment, Docker, git workflow patterns. ONLY durable rules — not transient operational events like "sandbox rebuilt" or "background task exited"
 - project-knowledge: Architecture, stack, database, API patterns
 - none: Casual conversation, questions, code snippets, or anything not worth remembering
 
@@ -39,7 +39,7 @@ const BATCH_SYSTEM_PROMPT = `You are a memory classifier for a developer tool. Y
 - correction: User correcting a mistake ("no, use X not Y", "wrong, it should be...")
 - preference: User stating a clear, reusable directive about how they want things done going forward ("we use tabs", "always use TypeScript", "I prefer X"). Must be a rule that applies beyond this conversation. NOT observations, complaints, questions, debugging statements, or one-off instructions like "fix this" or "tell me about X"
 - failure: Something broke or failed ("build failed", "error in deployment")
-- devops: CI/CD, deployment, Docker, git workflow patterns
+- devops: CI/CD, deployment, Docker, git workflow patterns. ONLY durable rules — not transient operational events like "sandbox rebuilt" or "background task exited"
 - project-knowledge: Architecture, stack, database, API patterns
 - none: Casual conversation, questions, code snippets, or anything not worth remembering
 
@@ -149,26 +149,36 @@ async function extractHindsightHint(failureDescription, context) {
         return null;
     }
 }
-const SESSION_EXTRACTION_PROMPT = `You are analyzing a coding session transcript to extract durable project knowledge.
+const SESSION_EXTRACTION_PROMPT = `You are analyzing a coding session transcript to extract durable lessons.
 
-The transcript shows tool calls (Bash, Edit, Read, Grep, etc.) and their results, plus user and assistant messages.
+The transcript shows tool calls (Bash, Edit, Read, Grep, etc.) and their results, plus user and assistant messages. Your primary job is to identify CAUSE-AND-EFFECT patterns — what failed, why, and what fixed it.
 
-Extract ONLY facts useful in FUTURE sessions:
-- Project conventions discovered (file structure, naming patterns, build tools, test frameworks)
-- Workflow patterns that worked or failed (e.g. "tests must be run from project root")
-- Technical constraints or gotchas encountered (e.g. "this project uses ESM, not CJS")
-- Environment requirements discovered (e.g. "needs Node 20+", "uses pnpm not npm")
+PRIORITY 1 — Failure → Fix sequences:
+Look for tool calls that failed (errors, timeouts, non-zero exits) followed by a different approach that succeeded. Extract the lesson as an imperative rule.
+Examples:
+- Command timed out because of interactive prompt → "scripts/upgrade-sandbox.sh requires interactive confirmation — pipe 'y' to auto-confirm"
+- Edit failed with old_string not found → "File X uses tabs not spaces — match exact indentation when editing"
+- Build failed after dependency change → "Run npm install after modifying package.json before building"
+
+PRIORITY 2 — Project conventions discovered:
+- File structure, naming patterns, build tools, test frameworks
+- Workflow patterns (e.g. "tests must be run from project root")
+- Technical constraints or gotchas (e.g. "this project uses ESM, not CJS")
+- Environment requirements (e.g. "needs Node 20+", "uses pnpm not npm")
+
+PRIORITY 3 — User corrections applied:
+- When the user corrected the agent's approach mid-session
 
 Do NOT extract:
-- Task-specific details (what was built, which files changed this session)
-- Debugging steps unlikely to recur
-- Code patterns visible by reading the codebase
+- Task-specific details (what was built, which files changed)
+- One-off debugging steps unlikely to recur
+- Code patterns derivable from reading the codebase
 - Anything in the EXISTING MEMORIES list below
 
 Respond with ONLY valid JSON (no markdown fences):
 [{"type":"project-knowledge|preference|devops|failure","content":"<imperative statement>","confidence":0.0-1.0}]
 
-Return [] if nothing durable was learned. Max 10 items. Each content should be a concise imperative statement (e.g. "Run tests with pnpm test, not npm test").`;
+Return [] if nothing durable was learned. Max 10 items. Each content should be a concise, actionable rule (e.g. "Pipe 'y' to scripts/upgrade-sandbox.sh — it has an interactive confirmation prompt").`;
 /**
  * Extract durable session learnings from a conversation summary using Haiku.
  * Returns null if no API key or on any failure.

package/dist/hooks/memory-stop-hook.js

@@ -94,8 +94,8 @@ async function handleMemoryStop(input) {
         // Reject extracts that are too short or too long (not clean rules)
         if (result.extract.length < 10 || result.extract.length > 200)
             continue;
-        // Corrections and preferences need higher confidence
-        if ((result.type === 'correction' || result.type === 'preference') && result.confidence < 0.7)
+        // Corrections, preferences, and devops need higher confidence to prevent noise
+        if ((result.type === 'correction' || result.type === 'preference' || result.type === 'devops') && result.confidence < 0.75)
             continue;
         if (result.confidence < 0.6)
             continue;

package/dist/hooks/tool-outcome-watcher.js

@@ -235,7 +235,10 @@ async function handleBashFailure(command, exitCode, output, sessionId) {
     const memoryService = memory_1.MemoryService.getInstance();
     memoryService.store({
         key,
-        value: failureContent,
+        value: {
+            content: `${failureContent.what_failed}. ${failureContent.why_failed}`,
+            ...failureContent,
+        },
         type: 'failure',
         context: { timestamp: Date.now() },
         relevanceScore: 0.85,

package/dist/shared/event-processors.js

@@ -206,7 +206,10 @@ function storeToolFailure(toolName, toolInput, output, sessionId) {
     const key = `hook_failure_${Date.now()}_${Math.random().toString(36).substr(2, 9)}`;
     memory_1.MemoryService.getInstance().store({
         key,
-        value: failureContent,
+        value: {
+            content: `${failureContent.what_failed}. ${failureContent.why_failed}`,
+            ...failureContent,
+        },
         type: 'failure',
         context: { timestamp: Date.now() },
         relevanceScore: 0.75,
@@ -278,7 +281,7 @@ async function processUserInput(text, sessionId) {
         return null;
     if (result.extract.length < 10 || result.extract.length > 200)
         return null;
-    if ((result.type === 'correction' || result.type === 'preference') && result.confidence < 0.7)
+    if ((result.type === 'correction' || result.type === 'preference' || result.type === 'devops') && result.confidence < 0.75)
         return null;
     if (result.confidence < 0.6)
         return null;
@@ -325,7 +328,7 @@ async function processSessionEnd(userTexts, sessionId, projectId, maxStore = 3)
                     continue;
                 if (result.extract.length < 10 || result.extract.length > 200)
                     continue;
-                if ((result.type === 'correction' || result.type === 'preference') && result.confidence < 0.7)
+                if ((result.type === 'correction' || result.type === 'preference' || result.type === 'devops') && result.confidence < 0.75)
                     continue;
                 if (result.confidence < 0.6)
                     continue;

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "claude-recall",
-  "version": "0.20.10",
+  "version": "0.20.12",
   "description": "Persistent memory for Claude Code and Pi with native Skills integration, automatic capture, failure learning, and project scoping",
   "main": "dist/index.js",
   "bin": {