owlmem@0.9.4
OwlMem command line: connect your AI apps to OwlMem in one command (npx -y owlmem install), an end-to-end encrypted vault for your AI agents, and a local MCP server.
AI Security Review
scanned 4d ago · by lpm-firewall-aiThe package’s auto-wake feature periodically reads local AI session logs and sends aggregated usage metrics to OwlMem, but the README does not disclose this reporting or describe a way to disable it separately. The reviewed flows do not show transcript text or credentials being sent to an unrelated host, so this warrants a telemetry warning rather than a malicious verdict.
Static reason
High-risk behavior combination matched malicious policy.
Trigger
Starts when the user enables auto-wake; the first usage report is scheduled after 30 seconds, then repeated periodically.
Impact
OwlMem receives information about local AI usage volume and quota status. The inspected report flow sends aggregates, not transcript text.
Mechanism
Reads local Claude Code and Codex JSONL session logs, derives aggregate token counts and usage status, and submits them to OwlMem. If the owner has enabled Claude quota reading, it also queries Anthropic with the local Claude OAuth token.
AI rationale
The package’s auto-wake feature periodically reads local AI session logs and sends aggregated usage metrics to OwlMem, but the README does not disclose this reporting or describe a way to disable it separately. The reviewed flows do not show transcript text or credentials being sent to an unrelated host, so this warrants a telemetry warning rather than a malicious verdict.
Evidence
owlmem.mjsREADME.md~/.claude/projects/**/*.jsonl~/.codex/sessions/**/*.jsonl~/.config/owlmem/wake/usage.json~/.claude/.credentials.jsonmacOS Keychain service Claude Code-credentials
Network endpoints2
owlmem.comapi.anthropic.com/api/oauth/usage
Decision evidence
public snapshotAI called this Suspicious at 91.0% confidence as Dangerous Capability with medium false-positive risk.
Evidence for warning
- Auto-wake scans Claude Code and Codex session transcript files on this machine.
- The scan derives daily input, output, and cache token counts from transcripts and sends aggregated usage and status data to OwlMem.
- The usage report runs after auto-wake starts and repeats on a timer.
- Claude quota lookup is gated by the server-provided `read_quota` setting and sends the Claude token to Anthropic’s own usage endpoint.
Evidence against
- The README presents auto-wake as an explicit feature users enable; it is not an install-time action.
- The report contains aggregated usage and quota status rather than transcript text.
- The Claude OAuth token is sent to `api.anthropic.com`, its own service, and only when quota reading is enabled.
Behavioral surface
ChildProcessCryptoEnvironmentVarsFilesystemNetworkShell
HighEntropyStringsUrlStrings
Source & flagged code
10 flaggedowlmem.mjsView file
14// owlmem vault ls
L15: // owlmem vault get <name> [field] prints the item (or one field) to stdout
L16: // owlmem vault put <name> [--desc text] < value-or-json
...
L37: import { readFileSync, writeFileSync, mkdirSync, existsSync, chmodSync, copyFileSync, rmSync, statSync, realpathSync, openSync, closeSync, appendFileSync, renameSync, readdirSync, ...
L38: import { gunzipSync } from "node:zlib";
L39: import { homedir, hostname, userInfo } from "node:os";
...
L41: import { fileURLToPath } from "node:url";
L42: import { spawn, spawnSync } from "node:child_process";
L43:
L44: export const VERSION = "0.9.4";
L45: const BASE = (process.env.OWLMEM_URL || "https://owlmem.com").replace(/\/$/, "");
L46: const DIR = process.env.OWLMEM_HOME || join(process.env.XDG_CONFIG_HOME || joi
Critical
Credential Exfiltration
Source appears to send environment or credential material to an external endpoint.
owlmem.mjsView on unpkg · L1414Trigger-reachable chain: manifest.bin -> owlmem.mjs
L14: // owlmem vault ls
L15: // owlmem vault get <name> [field] prints the item (or one field) to stdout
L16: // owlmem vault put <name> [--desc text] < value-or-json
...
L37: import { readFileSync, writeFileSync, mkdirSync, existsSync, chmodSync, copyFileSync, rmSync, statSync, realpathSync, openSync, closeSync, appendFileSync, renameSync, readdirSync, ...
L38: import { gunzipSync } from "node:zlib";
L39: import { homedir, hostname, userInfo } from "node:os";
...
L41: import { fileURLToPath } from "node:url";
L42: import { spawn, spawnSync } from "node:child_process";
L43:
L44: export const VERSION = "0.9.4";
L45: const BASE = (process.env.OWLMEM_URL || "https://owlmem.com").replace(/\/$/, "");
L46: const DIR = process.env.OW
Findings
2 Critical8 High3 Medium3 Low
CriticalCredential Exfiltrationowlmem.mjs
CriticalTrigger Reachable Dangerous Capabilityowlmem.mjs
HighChild Processowlmem.mjs
HighShellowlmem.mjs
HighSame File Env Network Executionowlmem.mjs
HighCommand Output Exfiltrationowlmem.mjs
HighSandbox Evasion Gated Capabilityowlmem.mjs
HighTrigger Reachable Credential Exfiltrationowlmem.mjs
HighTrigger Reachable Command Output Exfiltrationowlmem.mjs
HighSemantic Analysis Limitedowlmem.mjs
MediumNetwork
MediumEnvironment Vars
MediumStructural Risk Force Deep Review
Affected versions and remediation
This report applies to owlmem@0.9.4.
See all versions of owlmem for other recorded verdicts.
- Review the evidence and your use of owlmem@0.9.4 before allowing it. Restrict the permissions described in this report.
- Choose an independently verified alternative or release. This report does not establish that other versions are safe.
Evidence last updated: .
Think this verdict is wrong? Dispute it on GitHub or read how reports are produced.