Agent Economy ReportAgent Economy ReportCAUTION version 1.2.3 · audited 2026-09-17
This skill provides instructions and a small Node.js command-line tool for keeping AI agent memory in local Markdown files, with optional LanceDB, git-notes, SuperMemory, and Mem0 integrations. The bundled JavaScript only creates local memory files and reports their status; it does not make network requests, run shell commands, or harvest credentials. The main concerns are a destructive rm -rf command in the maintenance documentation, a recommendation to store an API key in ~/.zshrc, and optional third-party services that can receive memory content. No hidden prompt injection or credential exfiltration was found. Verdict: CAUTION.
| Declared purpose | Ultimate AI agent memory system for Cursor, Claude, ChatGPT & Copilot. WAL protocol, vector search, git-based knowledge graphs, cloud backup. Never lose context again. |
| Observed behavior | The package consists of SKILL.md, README.md, skill-card.md, package.json, and a Node.js CLI at bin/elite-memory.js. The CLI initializes SESSION-STATE.md, MEMORY.md, a memory/ directory, and dated daily log files in the current workspace, and checks their status using local filesystem APIs. It also checks for a LanceDB directory under the user's HOME directory. The documentation instructs the agent to run shell commands for git-notes, LanceDB, Mem0, and SuperMemory, and to export API keys. The bundled JavaScript contains no network calls, no shell execution, no download-and-execute behavior, and no obfuscation. |
| Verdict, rules only | CAUTION · reads credential-like environment variables: MEM0_API_KEY, OPENAI_API_KEY; touches persistence (cron, shell profile, autostart); uses sudo, rm -rf or dd |
| Verdict, AI | CAUTION |
| Final verdict | CAUTION (never better than either pass) |
| capability | vs purpose | detail |
|---|---|---|
| filesystem_write | needed | The CLI creates SESSION-STATE.md, MEMORY.md, a memory/ directory, and dated daily logs in the current workspace (bin/elite-memory.js lines 71, 79, 87, 98, 119). |
| filesystem_read | needed | The status command checks for and reads SESSION-STATE.md, MEMORY.md, and memory/ daily logs, and checks a LanceDB path under $HOME (bin/elite-memory.js lines 130-148 and 155). |
| shell | needed | Documentation tells the agent to run shell commands such as python3 memory.py, git init, mkdir, npm/npx, export API keys, and rm -rf (SKILL.md lines 100, 218, 228, 293, 334; README.md line 40). |
| network | needed | Optional Mem0 and SuperMemory integrations use third-party network APIs; the documentation also links to external sites. The bundled CLI itself makes no network calls. |
| credentials | needed | SKILL.md frontmatter requires OPENAI_API_KEY, and the documentation uses MEM0_API_KEY and SUPERMEMORY_API_KEY for optional cloud features. |
| install_packages | needed | package.json declares mem0ai as an optional dependency, and the documentation instructs users to run npm install mem0ai and npx elite-longterm-memory. |
| persistence | exceeds purpose | Documentation suggests adding SUPERMEMORY_API_KEY to ~/.zshrc for persistence, which stores a secret in a shell startup file. |
| severity | finding |
|---|---|
| high | Destructive recursive delete command The maintenance instructions include a recursive force delete of the LanceDB memory directory. If executed by an agent, this erases vector memory. It is documented as a 'nuclear option', but a careful user should review it before relying on agent automation. SKILL.md:293 rm -rf ~/.openclaw/memory/lancedb/ |
| medium | API key stored in shell startup file The optional SuperMemory setup suggests adding the SUPERMEMORY_API_KEY to ~/.zshrc. This stores a secret in plaintext in a shell startup file, making it available to all future shell sessions. SKILL.md:235 # Add to ~/.zshrc for persistence |
| low | Instruction prints secret to terminal The troubleshooting section tells the user to echo the OpenAI API key, which can expose the secret in terminal output or logs. SKILL.md:394 → Check OpenAI API key: `echo $OPENAI_API_KEY` |
| medium | Floating third-party dependency package.json declares mem0ai as an optional dependency using a caret version range. Install may pull newer minor versions, and the code of that third-party package is not reviewed here. package.json:36 "mem0ai": "^1.0.0" |
| medium | Conversation content can be sent to Mem0 The Mem0 example sends user content to the MemoryClient API. If enabled, conversation details leave the local environment and are processed by a third-party service. SKILL.md:344 { role: "user", content: "I prefer Tailwind over vanilla CSS" } |
| info | Agent workflow instruction The skill instructs the agent to write user-provided details to a local file before responding. This is consistent with the declared memory workflow, but it changes agent behavior and should be reviewed. SKILL.md:246 1. **User gives concrete detail?** → Write to SESSION-STATE.md BEFORE responding |
| info | Reads HOME environment variable The status command constructs a path from the HOME environment variable to check for LanceDB data. It does not read or transmit secrets. bin/elite-memory.js:155 const lancedbPath = path.join(process.env.HOME, '.clawdbot/memory/lancedb'); |
0 AI finding(s) were dropped because their file, line or quote did not match the package.
| domain | service rating | where |
|---|---|---|
| clawdhub.com | not in the service index | README.md:157 |
| clawhub.ai | not in the service index | skill-card.md:9 |
| github.com | not in the service index | README.md:158 |
| img.shields.io | not in the service index | README.md:5 |
| opensource.org | not in the service index | README.md:7 |
| www.npmjs.com | not in the service index | README.md:5 |
| x.com | not in the service index | README.md:162 |
| Binaries invoked | bash, git, node, npm, npx, python3, rm -rf |
| Environment variables read | DIR, HOME, MEM0_API_KEY, OPENAI_API_KEY |
| Hard-coded wallets | none |
| Pipes a download to a shell | 0 |
| eval / exec / subprocess | 0 |
| base64 blobs | 0 |
| File writes | 4 |
| Persistence | 1 |
| file | lines | sha256 |
|---|---|---|
README.md | 162 | 17c0c112c1363fe9… |
SKILL.md | 409 | 56bb341604f95198… |
_meta.json | 6 | 9c597953e49eab9b… |
bin/elite-memory.js | 187 | a937f5aab36c4b6e… |
package.json | 57 | 188612db5a262a29… |
skill-card.md | 63 | 94dccdf7f8226ddf… |
For agents
JSON: https://agenteconomy.report/k/elite-longterm-memory.audit.json · badge: https://agenteconomy.report/k/elite-longterm-memory.audited.svg ·
skill rating: /k/elite-longterm-memory · commission an audit of any skill: US$ 29 per version.
The complete published package of this exact version was downloaded from the registry and read statically; nothing was executed. A deterministic pass extracts network destinations, binaries, environment variables, writes, install commands, obfuscation markers, subprocesses and wallets, each with file and line. An AI then reads the whole package with those facts and writes the summary, the capabilities and the findings under a strict schema; every finding it produces must cite a file, a line and the exact text, or it is dropped. The final verdict is the worst of the two passes. The audit does not change the skill's trust tier (policy); the author may respond through the dispute channel and the response is published here. Commissioned by: the Agent Economy Report (free program: the 150 most downloaded skills, September 2026).