claude-recall 0.20.12 → 0.20.14

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

@@ -8,16 +8,23 @@ source: claude-recall
 
 # Corrections
 
-Auto-generated from 7 memories. Last updated: 2026-04-06.
+Auto-generated from 14 memories. Last updated: 2026-04-09.
 
 ## 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: 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: wait until next system performnce review
+- CORRECTION: [PreCompact] i just want you tellme what you observe claude recall has done lately
 - 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: [PreCompact] delete without asking first
 - CORRECTION: delete without asking first
 
 ---

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

@@ -1,15 +1,22 @@
 {
   "topicId": "corrections",
-  "sourceHash": "c3bf72dbbe482ded411952a50d3c0d6bb04f7c0a8d8ee76c01992ed8f27c8fcd",
-  "memoryCount": 7,
-  "generatedAt": "2026-04-06T21:44:03.092Z",
+  "sourceHash": "c32cc0fe5f8e375075a893c5c677945df13798f9aeda6763e5ee0a4b8de91a47",
+  "memoryCount": 14,
+  "generatedAt": "2026-04-09T20:39:45.990Z",
   "memoryKeys": [
+    "memory_1775767185973_5h07cll2t",
+    "memory_1775766122931_keod4qo79",
+    "memory_1775766107107_jh925r7qo",
+    "memory_1775766089878_07ouaa5vy",
+    "memory_1775766072071_snw3h6dke",
     "memory_1775511843079_fzzjymikj",
     "memory_1775511566528_2zqlbeiyy",
     "memory_1775500740471_dp5hmdxaf",
-    "memory_1775499328353_2ade6746o",
+    "hook_correction_1775554504739_geen26iio",
+    "hook_correction_1775543706749_u29qqfmdd",
     "hook_correction_1775511596163_oj20xa10q",
     "hook_correction_1775498361284_9amjuw4le",
+    "hook_correction_1775543706706_eqk8p0lm1",
     "hook_correction_1775511572913_yq1y9tvlo"
   ]
 }

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

@@ -8,18 +8,28 @@ source: claude-recall
 
 # Failure Lessons
 
-Auto-generated from 87 memories. Last updated: 2026-04-06.
+Auto-generated from 117 memories. Last updated: 2026-04-09.
 
 ## Rules
 
 - Check command syntax, file paths, and prerequisites before running
 - Check inputs and prerequisites before retrying
+- Multiple ssh tunnels + python processes writing same output file causes interleaved corruption and duplicate events. Root cause: nohup orphans from previous session not cleaned before restart.
 - 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
+- nemoclaw's SSRF guard rejects all private/loopback IPs — cannot use ericai-on-host locally without infrastructure patching at three layers; prefer public HTTPS API endpoints instead.
+- Migrating a 'messy and complicated application' as-is risks baking in workarounds, race conditions (e.g., two-writer reload-on-tick), and stale cache patterns (e.g., 5s cache) that no longer make sense in the new architecture — pause to identify and remove technical debt before migration
+- remediate-cell.js still uses legacy direct file mutation (path.join to mock-eiap/world-state.json). When run in sandbox, this fails — must refactor to call worldState.save() + strategic logging at caller boundaries, same pattern as remediate-backhaul.js.
 - 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.
+- Browser-SSO tokens are a fundamental impedance mismatch with headless workloads in K8s — do not attempt to make Kiro (browser-token-based auth) work in containerized environments; use API tokens or service accounts instead
+- Avoid: Command failed: openshell --help 2>&1 | grep -iE "gateway|stop|start|down" | head -15; echo ---gateway-list---; o... → Instead: Check command syntax, file paths, and prerequisites before running
+- Avoid: Command failed: echo ---kill-everything---; pkill -f "stream-sessions.sh" 2>/dev/null; pkill -f "stream-watcher.p... → Instead: Check command syntax, file paths, and prerequisites before running
+- Avoid: Command failed: docker exec openshell-cluster-nemoclaw kubectl exec -n openshell nka-telco -- sh -c "ls -la /sand... → Instead: Check command syntax, file paths, and prerequisites before running
+- Avoid: Command failed: cd /home/ebiarao/.claude/projects/-home-ebiarao-repos-wsl-personal-projects-openclaw-autonomous-t... → Instead: Check command syntax, file paths, and prerequisites before running
+- Avoid: Command failed: B64=$(base64 -w0 < agents/nka/scripts/remediate-backhaul.js); docker exec openshell-cluster-nemoc... → Instead: Check command syntax, file paths, and prerequisites before running
 - Avoid: Command failed: git commit -m "$(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
@@ -50,6 +60,22 @@ Event-driven agent cascade + direct EIAP polling — two breakthroughs... → In
 - 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: sqlite3 ~/.claude-recall/claude-recall.db \"SELECT DISTINCT content FROM memories WHERE content LI...","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 content (17014 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 (17014 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":"Grep failed","why_failed":"spawn /home/ebiarao/.nvm/versions/node/v20.19.3/lib/node_modules/@anthropic-ai/claude-code/vendor/ripgrep/x64-linux/rg ENOENT","what_should_do":"Check inputs and prerequisites before retrying","context":"Error: spawn /home/ebiarao/.nvm/versions/node/v20.19.3/lib/node_modules/@anthropic-ai/claude-code/vendor/ripgrep/x64-linux/rg ENOENT","preventative_checks":["Verify tool inputs are correct","Check preconditions"]}
+- {"what_failed":"Read failed","why_failed":"File content (17014 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 (17014 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: tail -30 ~/.claude-recall/hook-logs/event-processor.log ~/.claude-recall/hook-logs/memory-stop.lo...","why_failed":"Exit code 1\ntail: option used in invalid context -- 3","what_should_do":"Check inputs and prerequisites before retrying","context":"Bash error: Exit code 1\ntail: option used in invalid context -- 3","preventative_checks":["Verify tool inputs are correct","Check preconditions"]}
+- {"what_failed":"Bash command failed: npx claude-recall outcomes 2>&1","why_failed":"Exit code 1\nerror: unknown command 'outcomes'","what_should_do":"Check inputs and prerequisites before retrying","context":"Bash error: Exit code 1\nerror: unknown command 'outcomes'","preventative_checks":["Verify tool inputs are correct","Check preconditions"]}
+- Read/ToolSearch hook errors in claude-recall indicate search_enforcer.py rule-load failures — check python3 syntax and MCP tool availability (mcp__claude-recall__mcp__claude-recall__load_rules must be registered).
+- Pre-tool-use hook errors (e.g., '[python3 search_enforcer.py]' load failures) indicate stale or misconfigured rule state. Call mcp__claude-recall__load_rules {} before retrying the tool to refresh rule context.
+- {"what_failed":"Read failed","why_failed":"File content (11705 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 (11705 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"]}
+- git mv may not fully track deletion of nested files — after moving directories, verify with git status that all expected deletes are staged; 'git mv mock-eiap/ sandbox/mock-eiap/' leaves orphaned nested files (data.js, world/build-ireland.js, world/config.js) that must be explicitly deleted
+- {"what_failed":"Bash command failed: echo ---kill-smoke-eiap---; pkill -f \"mock-eiap/server.js\" 2>/dev/null; sleep 1; echo ---try-even...","why_failed":"Exit code 144\n---kill-smoke-eiap---","what_should_do":"Check inputs and prerequisites before retrying","context":"Bash error: Exit code 144\n---kill-smoke-eiap---","preventative_checks":["Verify tool inputs are correct","Check preconditions"]}
+- {"what_failed":"Bash command failed: openshell --help 2>&1 | grep -iE \"gateway|stop|start|down\" | head -15; echo ---gateway-list---; o...","why_failed":"Exit code 2\nGATEWAY COMMANDS\n  gateway:     Manage the gateway lifecycle\n  status:      Show gateway status and information\n  doctor:      Diagnose gateway issues\n  -g, --gateway <GATEWAY>\n        ...","what_should_do":"Check inputs and prerequisites before retrying","context":"Bash error: Exit code 2\nGATEWAY COMMANDS\n  gateway:     Manage the gateway lifecycle\n  status:      Show gateway status and information\n  doctor:      Diagnose gateway issues\n  -g, --gateway <GATEWAY>\n        ...","preventative_checks":["Verify tool inputs are correct","Check preconditions"]}
+- {"what_failed":"Bash command failed: echo ---kill-everything---; pkill -f \"stream-sessions.sh\" 2>/dev/null; pkill -f \"stream-watcher.p...","why_failed":"Exit code 144\n---kill-everything---","what_should_do":"Check inputs and prerequisites before retrying","context":"Bash error: Exit code 144\n---kill-everything---","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 -- sh -c \"ls -la /sand...","why_failed":"Exit code 2\nls: cannot access '/sandbox/.openclaw-data/workspace/artifacts/rebuild-status.json': No such file or directory\ncat: /sandbox/.openclaw-data/workspace/artifacts/rebuild-status.json: No s...","what_should_do":"Check inputs and prerequisites before retrying","context":"Bash error: Exit code 2\nls: cannot access '/sandbox/.openclaw-data/workspace/artifacts/rebuild-status.json': No such file or directory\ncat: /sandbox/.openclaw-data/workspace/artifacts/rebuild-status.json: No s...","preventative_checks":["Verify tool inputs are correct","Check preconditions"]}
+- {"what_failed":"Bash command failed: cd /home/ebiarao/.claude/projects/-home-ebiarao-repos-wsl-personal-projects-openclaw-autonomous-t...","why_failed":"Exit code 2\nls: cannot access 'project_open_remediate_cell_rewrite.md': No such file or directory\nproject_openclaw_egress_filter.md","what_should_do":"Check inputs and prerequisites before retrying","context":"Bash error: Exit code 2\nls: cannot access 'project_open_remediate_cell_rewrite.md': No such file or directory\nproject_openclaw_egress_filter.md","preventative_checks":["Verify tool inputs are correct","Check preconditions"]}
+- {"what_failed":"Bash command failed: B64=$(base64 -w0 < agents/nka/scripts/remediate-backhaul.js); docker exec openshell-cluster-nemoc...","why_failed":"Exit code 2\nsh: 1: cannot create /sandbox/agents/nka/scripts/remediate-backhaul.js: Directory nonexistent\ncommand terminated with exit code 2","what_should_do":"Check inputs and prerequisites before retrying","context":"Bash error: Exit code 2\nsh: 1: cannot create /sandbox/agents/nka/scripts/remediate-backhaul.js: Directory nonexistent\ncommand terminated with exit code 2","preventative_checks":["Verify tool inputs are correct","Check preconditions"]}
 - {"what_failed":"Bash command failed: git commit -m \"$(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"]}
@@ -91,12 +117,16 @@ Event-driven agent cascade + direct EIAP polling — two breakthroughs... → In
 - {"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
+- 'npx claude-recall stats' can fail with 'PreToolUse:Bash hook error' — indicates a disconnect between hook registration and execution. When stats fails, verify search results separately as they may disagree; at least one is silently wrong.
 - 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
+- Bash command base64 encoding long file content failed silently — use base64 -w0 to disable line wrapping and pipe directly into kubectl exec without intermediate variables
+- Read hook error on .gitignore — search_enforcer.py failed to load rules; retry Read with mcp__claude-recall__load_rules first to populate rules before file operations
+- Avoid: Claude backtracked: "Going back to" → Instead: Consider the problem more carefully before starting. Check for similar past failures.
 - 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

@@ -1,17 +1,27 @@
 {
   "topicId": "failure-lessons",
-  "sourceHash": "0a433bc57da8ed5c3137f8676cae1ec253e4a5037b115400d0e0c9e394789440",
-  "memoryCount": 87,
-  "generatedAt": "2026-04-06T21:44:03.038Z",
+  "sourceHash": "f5b6a2ae08e9e7fbfe9f9d678b1068fa5eaad0d81bf30ef2e3d065bfcf873626",
+  "memoryCount": 117,
+  "generatedAt": "2026-04-09T20:21:12.032Z",
   "memoryKeys": [
     "promoted_1775498319002_awwrsovw6",
     "promoted_1775499579486_bxvgjc80r",
+    "hook_failure_1775567919637_ado9z7i52",
     "hook_failure_1775502039213_k4yusuan1",
     "hook_failure_1772637584921_0tj4rrxnt",
     "hook_failure_1775503464385_vz5vdzxvs",
     "hook_failure_1775500959351_j74owpvko",
+    "hook_failure_1775637261359_prhnbz9gp",
+    "hook_failure_1775571333207_b0jxc7wg9",
+    "hook_failure_1775554467902_vefk0wqnx",
     "hook_failure_1775511109813_ly4cjtde6",
     "hook_failure_1775509498191_f79imbrs4",
+    "hook_failure_1775645996326_0mz9beu5u",
+    "hook_failure_non-zero-exit_1775568947522_v9v3s4rcq",
+    "hook_failure_non-zero-exit_1775567518525_sa9trys9z",
+    "hook_failure_non-zero-exit_1775566023853_x9yadg72p",
+    "hook_failure_non-zero-exit_1775564413767_rjg38weet",
+    "hook_failure_non-zero-exit_1775546352476_ifdye1dp6",
     "hook_failure_non-zero-exit_1775511575363_nen5eyl3b",
     "hook_failure_non-zero-exit_1775510785418_2xm52g5q8",
     "hook_failure_non-zero-exit_1775510309712_olv4egt1g",
@@ -41,6 +51,22 @@
     "hook_failure_non-zero-exit_1772640279977_g3gwlfoqi",
     "hook_failure_1772637570984_yxs8zmurp",
     "hook_failure_1772637485532_djep8eysa",
+    "hook_failure_1775764137796_nt2gy4gn4",
+    "hook_failure_1775764050725_qexdbw5km",
+    "hook_failure_1775764035582_dbcp1mjyw",
+    "hook_failure_1775764006133_rszt1zn54",
+    "hook_failure_1775763649047_jnx0je5mk",
+    "hook_failure_1775763633851_mga29nvnf",
+    "hook_failure_1775760197556_7kmkzjbph",
+    "hook_failure_1775759374123_6l7pw3gqs",
+    "hook_failure_1775759306394_fm3fhkc35",
+    "hook_failure_1775571478609_zua292iy0",
+    "hook_failure_1775570162971_5ii1j8sw3",
+    "hook_failure_1775568820718_1d7pfm9s1",
+    "hook_failure_1775566682561_jfbkzxzvt",
+    "hook_failure_1775565590990_g2pab28yu",
+    "hook_failure_1775563474928_wqww7caqd",
+    "hook_failure_1775544637722_0vkaiyddz",
     "hook_failure_1775511332509_3mxf5vtg3",
     "hook_failure_1775510710817_sn23160zp",
     "hook_failure_1775510222535_o2t7vm7qp",
@@ -82,12 +108,16 @@
     "hook_failure_1775498542480_wc23ux8gt",
     "hook_failure_silent-test-failure_1772640279996_m8f3ks8fw",
     "hook_failure_1772637495154_i21dho3dv",
+    "hook_failure_1775760542139_lm1nf043v",
     "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_1775546352399_1znn01j7o",
+    "hook_failure_1775759908170_k8yreza2o",
+    "hook_failure_backtracking_1775564413782_4qgh3envk",
     "hook_failure_1775509857057_847nm61ld",
     "hook_failure_backtracking_1775508193524_54wn8o688",
     "hook_failure_backtracking_1772640280006_snb0x5cgg"

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

@@ -8,33 +8,15 @@ source: claude-recall
 
 # Preferences
 
-Auto-generated from 23 memories. Last updated: 2026-04-06.
+Auto-generated from 5 memories. Last updated: 2026-04-10.
 
 ## Rules
 
-- Session test preference 1775511843229
-- Test preference 1775511843109-2
-- Test preference 1775511843109-1
-- Test preference 1775511843109-0
+- Session test preference 1775821892450
+- Test preference 1775821892400-2
+- Test preference 1775821892400-1
+- Test preference 1775821892400-0
 - 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
-- 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.
-- 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.
-- 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
 
 ---
 *Auto-generated by Claude Recall. Regenerate: `npx claude-recall skills generate`*

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

@@ -1,31 +1,13 @@
 {
   "topicId": "preferences",
-  "sourceHash": "819c9fa966f5ffabe7b065fa6290958f1601072fb16e47d162bd513228ec08d5",
-  "memoryCount": 23,
-  "generatedAt": "2026-04-06T21:44:03.259Z",
+  "sourceHash": "9b2ae714e4bdd82091b4fcc5347c2b1a8a2a10d3be57ca57877e943e380a5457",
+  "memoryCount": 5,
+  "generatedAt": "2026-04-10T11:51:32.469Z",
   "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_1775208934902_2kovciriy",
-    "memory_1775208477621_fqa3w21j1",
-    "memory_1775168901303_o8o0i4952",
-    "hook_preference_1775511561137_rjm92h4o3",
-    "hook_preference_1775508193449_aphyh28z9",
-    "hook_preference_1775503672864_weqgjstda",
-    "hook_preference_1775498374200_6d21qj14y"
+    "memory_1775821892451_dzjdxx6ef",
+    "memory_1775821892427_462vxb2gq",
+    "memory_1775821892416_8taocju9a",
+    "memory_1775821892402_3x8ksma5p",
+    "memory_1775821892372_4khhv0wc2"
   ]
 }

package/dist/cli/claude-recall-cli.js

@@ -117,10 +117,9 @@ class ClaudeRecallCLI {
      * Get stats for a specific project (includes universal and unscoped memories)
      */
     getProjectStats(projectId) {
-        const allMemories = this.memoryService.search('');
-        const projectMemories = allMemories.filter(m => m.project_id === projectId ||
-            m.scope === 'universal' ||
-            m.project_id === null);
+        // Use getAllByProject for true enumeration. search() pre-ranks and caps
+        // at top-5, which silently masks most memories from stats.
+        const projectMemories = this.memoryService.getAllByProject(projectId);
         // Calculate byType breakdown
         const byType = {};
         for (const mem of projectMemories) {
@@ -135,7 +134,8 @@ class ClaudeRecallCLI {
      * Show skills evolution breakdown
      */
     showSkillsEvolution() {
-        const allMemories = this.memoryService.search('');
+        const projectId = config_1.ConfigService.getInstance().getProjectId();
+        const allMemories = this.memoryService.getAllByProject(projectId);
         const devopsMemories = allMemories.filter(m => m.type === 'devops');
         if (devopsMemories.length === 0) {
             return;
@@ -182,7 +182,8 @@ class ClaudeRecallCLI {
         // Rough estimation:
         // - Each memory saves ~200 tokens (vs repeating to LLM)
         // - DevOps memories save ~1,500 tokens each (vs loading reference files)
-        const allMemories = this.memoryService.search('');
+        const projectId = config_1.ConfigService.getInstance().getProjectId();
+        const allMemories = this.memoryService.getAllByProject(projectId);
         const devopsCount = allMemories.filter(m => m.type === 'devops').length;
         const otherCount = stats.total - devopsCount;
         const devopsSavings = devopsCount * 1500;
@@ -224,11 +225,9 @@ class ClaudeRecallCLI {
      * Show failure memories (v0.7.0)
      */
     showFailures(options) {
-        const allMemories = this.memoryService.search('');
-        let failures = allMemories.filter(m => m.type === 'failure');
-        if (options.project) {
-            failures = failures.filter(m => m.project_id === options.project);
-        }
+        const projectId = options.project || config_1.ConfigService.getInstance().getProjectId();
+        const allMemories = this.memoryService.getAllByProject(projectId);
+        const failures = allMemories.filter(m => m.type === 'failure');
         // Sort by timestamp (newest first)
         failures.sort((a, b) => (b.timestamp || 0) - (a.timestamp || 0));
         // Limit results
@@ -354,8 +353,8 @@ class ClaudeRecallCLI {
         // Determine search scope
         let results;
         if (options.global) {
-            // Global search: all memories
-            results = this.memoryService.search(query);
+            // Global search: explicit cross-project opt-in
+            results = this.memoryService.search(query, { includeAllProjects: true });
         }
         else if (options.project) {
             // Project-specific search: project + universal
@@ -408,8 +407,10 @@ class ClaudeRecallCLI {
     async export(outputPath, options) {
         const format = options.format || 'json';
         try {
-            // Get all memories via search with empty query
-            const memories = this.memoryService.search('');
+            // Export all memories from the current project (+ universal). For a
+            // true cross-project dump, the user can run `--project` per project.
+            const projectId = config_1.ConfigService.getInstance().getProjectId();
+            const memories = this.memoryService.getAllByProject(projectId);
             if (format === 'json') {
                 const exportData = {
                     version: '0.2.0',

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

@@ -49,6 +49,7 @@ var __importStar = (this && this.__importStar) || (function () {
 Object.defineProperty(exports, "__esModule", { value: true });
 exports.handleBashFailureWatcher = void 0;
 exports.handleToolOutcomeWatcher = handleToolOutcomeWatcher;
+exports.shouldCaptureFailure = shouldCaptureFailure;
 exports.handleToolFailure = handleToolFailure;
 const fs = __importStar(require("fs"));
 const path = __importStar(require("path"));
@@ -195,6 +196,37 @@ async function handleToolOutcomeWatcher(input) {
 }
 // Backward compatibility alias
 exports.handleBashFailureWatcher = handleToolOutcomeWatcher;
+/**
+ * Quality filter: should this bash failure be captured as a memory?
+ * Returns false for exploratory probes, command-not-found, and other low-signal failures.
+ */
+function shouldCaptureFailure(command, exitCode, output) {
+    // Very short commands (ls, cd, etc.) — not worth capturing
+    if (command.trim().length < 5)
+        return false;
+    // Exploratory probes: commands ending in 2>/dev/null are intentionally suppressing errors
+    if (/2>\s*\/dev\/null\s*$/.test(command))
+        return false;
+    // Command-existence checks: which, type, command -v
+    const trimmed = command.trim();
+    if (/^(which|type)\s+/i.test(trimmed))
+        return false;
+    if (/^command\s+-v\s+/i.test(trimmed))
+        return false;
+    // Command not found (exit code 127 or output contains "command not found")
+    if (exitCode === '127')
+        return false;
+    if (/command not found/i.test(output))
+        return false;
+    // No useful output — nothing to learn from
+    const cleanOutput = output.replace(/\(no output\)/gi, '').trim();
+    if (cleanOutput.length < 10 && !/^(npm|npx|node|python|pip|cargo|go|make|docker|git)\s/.test(trimmed))
+        return false;
+    // EISDIR / "Is a directory" — file read probes
+    if (/is a directory/i.test(output) || /EISDIR/i.test(output))
+        return false;
+    return true;
+}
 // --- Bash handler (original bash-failure-watcher logic) ---
 async function handleBashOutcome(input) {
     const command = input.tool_input?.command;
@@ -215,6 +247,11 @@ async function handleBashOutcome(input) {
     }
 }
 async function handleBashFailure(command, exitCode, output, sessionId) {
+    // Quality filter — skip low-signal failures
+    if (!shouldCaptureFailure(command, exitCode, output)) {
+        (0, shared_1.hookLog)(HOOK_NAME, `Skipped low-quality failure: ${truncate(command, 60)} (exit ${exitCode})`);
+        return;
+    }
     // Dedup check
     const existing = (0, shared_1.searchExisting)(command);
     if ((0, shared_1.isDuplicate)(command, existing, 0.7)) {

package/dist/mcp/prompts-handler.js

@@ -161,7 +161,7 @@ class PromptsHandler {
      * Injects all user preferences into context
      */
     async getPreferencesPrompt() {
-        const preferences = this.memoryStorage.searchByContext({ type: 'preference' });
+        const preferences = this.memoryStorage.searchByContext({ type: 'preference', project_id: config_1.ConfigService.getInstance().getProjectId() });
         const preferencesText = this.formatPreferencesForPrompt(preferences);
         return {
             description: 'User coding preferences automatically injected',
@@ -182,7 +182,7 @@ Apply these preferences when generating code or making suggestions.`
      * Injects relevant project knowledge
      */
     async getProjectContextPrompt(topic) {
-        let knowledge = this.memoryStorage.searchByContext({ type: 'project-knowledge' });
+        let knowledge = this.memoryStorage.searchByContext({ type: 'project-knowledge', project_id: config_1.ConfigService.getInstance().getProjectId() });
         // Filter by topic if provided
         if (topic) {
             const results = this.memoryService.search(topic);
@@ -211,7 +211,7 @@ Use this information when working with the project.`
      * Injects recent correction patterns
      */
     async getCorrectionsPrompt() {
-        const corrections = this.memoryStorage.searchByContext({ type: 'correction' });
+        const corrections = this.memoryStorage.searchByContext({ type: 'correction', project_id: config_1.ConfigService.getInstance().getProjectId() });
         // Get most recent 10
         const recentCorrections = corrections
             .sort((a, b) => (b.timestamp || 0) - (a.timestamp || 0))

package/dist/mcp/resources-handler.js

@@ -3,6 +3,7 @@ Object.defineProperty(exports, "__esModule", { value: true });
 exports.ResourcesHandler = void 0;
 const memory_1 = require("../services/memory");
 const logging_1 = require("../services/logging");
+const config_1 = require("../services/config");
 /**
  * Handles MCP Resources protocol
  * Resources are static/semi-static context that Claude Code can subscribe to
@@ -141,7 +142,7 @@ class ResourcesHandler {
      * Get preferences resource content
      */
     async getPreferencesResource(uri) {
-        const memories = this.memoryStorage.searchByContext({ type: 'preference' });
+        const memories = this.memoryStorage.searchByContext({ type: 'preference', project_id: config_1.ConfigService.getInstance().getProjectId() });
         // Format preferences for easy consumption
         const preferences = memories.map(m => ({
             key: m.key,
@@ -163,7 +164,7 @@ class ResourcesHandler {
      * Get project knowledge resource content
      */
     async getProjectKnowledgeResource(uri) {
-        const memories = this.memoryStorage.searchByContext({ type: 'project-knowledge' });
+        const memories = this.memoryStorage.searchByContext({ type: 'project-knowledge', project_id: config_1.ConfigService.getInstance().getProjectId() });
         const knowledge = memories.map(m => ({
             key: m.key,
             content: m.value,
@@ -182,7 +183,7 @@ class ResourcesHandler {
      * Get recent corrections resource content
      */
     async getCorrectionsResource(uri) {
-        const memories = this.memoryStorage.searchByContext({ type: 'correction' });
+        const memories = this.memoryStorage.searchByContext({ type: 'correction', project_id: config_1.ConfigService.getInstance().getProjectId() });
         // Get most recent 10 corrections
         const recentCorrections = memories
             .sort((a, b) => (b.timestamp || 0) - (a.timestamp || 0))
@@ -206,7 +207,7 @@ class ResourcesHandler {
     async getActiveContextResource(uri) {
         // Get top 5 most accessed memories from last 24 hours
         const oneDayAgo = Date.now() - (24 * 60 * 60 * 1000);
-        const allMemories = this.memoryStorage.searchByContext({});
+        const allMemories = this.memoryStorage.searchByContext({ project_id: config_1.ConfigService.getInstance().getProjectId() });
         const recentMemories = allMemories
             .filter(m => (m.timestamp || 0) > oneDayAgo)
             .sort((a, b) => (b.access_count || 0) - (a.access_count || 0))

package/dist/memory/pattern-store.js

@@ -36,7 +36,8 @@ class PatternStore {
         const memories = this.storage.searchByContext({
             type: 'correction-pattern',
             project_id: process.env.CLAUDE_PROJECT_DIR,
-            file_path: undefined
+            file_path: undefined,
+            includeAllProjects: !process.env.CLAUDE_PROJECT_DIR,
         });
         for (const memory of memories) {
             const stored = memory.value;
@@ -55,7 +56,8 @@ class PatternStore {
         const patterns = this.storage.searchByContext({
             type: 'correction-pattern',
             project_id: process.env.CLAUDE_PROJECT_DIR,
-            file_path: undefined
+            file_path: undefined,
+            includeAllProjects: !process.env.CLAUDE_PROJECT_DIR,
         });
         return patterns
             .map(m => m.value)

package/dist/memory/storage.js

@@ -230,6 +230,77 @@ class MemoryStorage {
         });
         return crypto.createHash('sha256').update(canonical).digest('hex');
     }
+    /**
+     * Find a same-type memory whose text content is a near-duplicate (Jaccard >= 0.85).
+     * Returns the key of the matching memory, or null if none found.
+     */
+    findFuzzyDuplicate(memory) {
+        const newText = this.extractText(memory.value).toLowerCase();
+        if (newText.length < 40)
+            return null; // Too short to fuzzy-match reliably
+        // Scope fuzzy dedup to same type AND same project (or both null/universal)
+        const projectFilter = memory.project_id
+            ? 'AND project_id = ?'
+            : 'AND (project_id IS NULL OR project_id = \'\')';
+        const params = [memory.type, memory.key];
+        if (memory.project_id)
+            params.push(memory.project_id);
+        const candidates = this.db.prepare(`SELECT key, value FROM memories WHERE type = ? AND key != ? ${projectFilter}`).all(...params);
+        for (const candidate of candidates) {
+            let candidateText;
+            try {
+                const parsed = JSON.parse(candidate.value);
+                candidateText = this.extractText(parsed).toLowerCase();
+            }
+            catch {
+                candidateText = candidate.value.toLowerCase();
+            }
+            if (this.jaccardSimilarity(newText, candidateText) >= 0.65) {
+                return candidate.key;
+            }
+        }
+        return null;
+    }
+    extractText(value) {
+        if (typeof value === 'string')
+            return value;
+        if (value && typeof value === 'object') {
+            // Collect all leaf string values, ignoring JSON keys
+            const leaves = [];
+            const collect = (obj) => {
+                if (typeof obj === 'string') {
+                    leaves.push(obj);
+                    return;
+                }
+                if (obj && typeof obj === 'object') {
+                    for (const v of Object.values(obj))
+                        collect(v);
+                }
+            };
+            collect(value);
+            return leaves.join(' ');
+        }
+        return String(value);
+    }
+    jaccardSimilarity(a, b) {
+        const tokenize = (s) => {
+            const words = s.replace(/[^a-z0-9\s]/g, ' ').split(/\s+/).filter(Boolean);
+            return new Set(words.filter(w => !MemoryStorage.STOP_WORDS.has(w)));
+        };
+        const wordsA = tokenize(a);
+        const wordsB = tokenize(b);
+        if (wordsA.size === 0 && wordsB.size === 0)
+            return 1;
+        if (wordsA.size === 0 || wordsB.size === 0)
+            return 0;
+        let intersection = 0;
+        for (const w of wordsA) {
+            if (wordsB.has(w))
+                intersection++;
+        }
+        const union = wordsA.size + wordsB.size - intersection;
+        return union === 0 ? 0 : intersection / union;
+    }
     save(memory) {
         const contentHash = this.computeContentHash(memory.value, memory.type);
         // Write-time dedup: check if identical content already exists under a different key
@@ -240,6 +311,13 @@ class MemoryStorage {
             this.db.pragma('wal_checkpoint(TRUNCATE)');
             return;
         }
+        // Fuzzy dedup: check if a same-type memory with very similar content exists
+        const fuzzyMatch = this.findFuzzyDuplicate(memory);
+        if (fuzzyMatch) {
+            this.db.prepare('UPDATE memories SET timestamp = ?, access_count = access_count + 1 WHERE key = ?').run(Date.now(), fuzzyMatch);
+            this.db.pragma('wal_checkpoint(TRUNCATE)');
+            return;
+        }
         const stmt = this.db.prepare(`
       INSERT OR REPLACE INTO memories
       (key, value, type, project_id, file_path, timestamp, relevance_score, access_count,
@@ -296,6 +374,13 @@ class MemoryStorage {
         };
     }
     searchByContext(context) {
+        // Fail loud when called without scope. This prevents silent cross-project
+        // leakage that the previous default-allow behaviour caused. Callers that
+        // genuinely want all projects must opt in with includeAllProjects: true.
+        if (!context.project_id && !context.includeAllProjects) {
+            throw new Error('searchByContext requires context.project_id or context.includeAllProjects=true. ' +
+                'Calling without either would silently return memories from all projects.');
+        }
         let query = 'SELECT * FROM memories WHERE 1=1';
         const params = [];
         if (context.project_id) {
@@ -303,6 +388,7 @@ class MemoryStorage {
             query += ' AND (project_id = ? OR scope = ? OR project_id IS NULL)';
             params.push(context.project_id, 'universal');
         }
+        // includeAllProjects: no project filter — caller explicitly asked for everything.
         if (context.file_path) {
             query += ' AND file_path = ?';
             params.push(context.file_path);
@@ -533,3 +619,9 @@ class MemoryStorage {
     }
 }
 exports.MemoryStorage = MemoryStorage;
+MemoryStorage.STOP_WORDS = new Set([
+    'a', 'an', 'the', 'is', 'are', 'was', 'were', 'be', 'been', 'being',
+    'to', 'of', 'in', 'for', 'on', 'with', 'at', 'by', 'from', 'as',
+    'and', 'or', 'but', 'not', 'no', 'do', 'does', 'did', 'this', 'that',
+    'it', 'its', 'via', 'can', 'should', 'will', 'would', 'may', 'might',
+]);

package/dist/services/config.js

@@ -134,12 +134,16 @@ class ConfigService {
         return { ...this.config };
     }
     getDatabasePath() {
-        // ALWAYS use ~/.claude-recall/claude-recall.db regardless of environment variables
-        // This ensures consistency across CLI and MCP server
-        const dbDir = path.join(os.homedir(), '.claude-recall');
-        const dbPath = path.join(dbDir, 'claude-recall.db');
-        // Ensure database directory exists
-        if (!fs.existsSync(dbDir)) {
+        // Honor CLAUDE_RECALL_DB_PATH / CLAUDE_RECALL_DB_NAME via the loaded
+        // config (loadConfig already reads them with sensible defaults). The
+        // previous "always use ~/.claude-recall" behaviour broke test isolation:
+        // tests setting CLAUDE_RECALL_DB_PATH silently wrote to the production
+        // DB and polluted it with fixture rows.
+        const dbDir = this.config.database.path;
+        const dbPath = path.join(dbDir, this.config.database.name);
+        // Ensure database directory exists. Skip for in-memory DBs (`:memory:`)
+        // and any path that doesn't look like a real filesystem location.
+        if (dbDir && dbDir !== ':memory:' && !fs.existsSync(dbDir)) {
             fs.mkdirSync(dbDir, { recursive: true });
         }
         return dbPath;

package/dist/services/memory.js

@@ -109,19 +109,35 @@ class MemoryService {
         }
     }
     /**
-     * Search memories by keyword
+     * Search memories by keyword.
+     *
+     * Scope rules (post-Fix-2):
+     *   - Default: scoped to current project (projectId from ConfigService) +
+     *     universal/unscoped memories.
+     *   - Pass opts.projectId to scope to a specific project.
+     *   - Pass opts.includeAllProjects=true to opt into a true global search
+     *     (used by `claude-recall search --global`).
      */
-    search(query, sortBy = 'relevance') {
+    search(query, optsOrSortBy = 'relevance') {
+        const opts = typeof optsOrSortBy === 'string' ? { sortBy: optsOrSortBy } : optsOrSortBy;
+        const sortBy = opts.sortBy || 'relevance';
         try {
-            // Use findRelevant with query context for better semantic matching
             const context = {
                 query: query,
-                timestamp: Date.now()
+                timestamp: Date.now(),
             };
+            if (opts.includeAllProjects) {
+                context.includeAllProjects = true;
+            }
+            else {
+                context.project_id = opts.projectId || this.config.getProjectId();
+            }
             const results = this.retrieval.findRelevant(context, sortBy);
             this.logger.logRetrieval(query, results.length, {
                 searchType: 'contextual',
-                sortBy
+                sortBy,
+                projectId: context.project_id,
+                includeAllProjects: !!opts.includeAllProjects,
             });
             return results;
         }
@@ -130,6 +146,24 @@ class MemoryService {
             throw error;
         }
     }
+    /**
+     * Enumerate ALL memories scoped to a project (project + universal + unscoped),
+     * with NO ranking and NO result cap.
+     *
+     * Use this for stats, exports, and any other "give me everything in scope"
+     * operation. Do NOT use search() / findRelevant() for enumeration — those
+     * pre-rank by type priority and cap at top-5, which silently hides most
+     * memories from callers that wanted a complete list.
+     */
+    getAllByProject(projectId) {
+        try {
+            return this.storage.searchByContext({ project_id: projectId });
+        }
+        catch (error) {
+            this.logger.logServiceError('MemoryService', 'getAllByProject', error, { projectId });
+            throw error;
+        }
+    }
     /**
      * Get memory storage statistics
      */

package/dist/services/skill-generator.js

@@ -288,6 +288,10 @@ class SkillGenerator {
         if (projectId) {
             searchContext.project_id = projectId;
         }
+        else {
+            // Skill generation runs across all projects for global skill discovery.
+            searchContext.includeAllProjects = true;
+        }
         let memories = this.storage.searchByContext(searchContext);
         // Filter by devops category if specified
         if (topic.devopsCategory) {

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "claude-recall",
-  "version": "0.20.12",
+  "version": "0.20.14",
   "description": "Persistent memory for Claude Code and Pi with native Skills integration, automatic capture, failure learning, and project scoping",
   "main": "dist/index.js",
   "bin": {