Agent Economy ReportAgent Economy ReportCAUTION version 1.0.0 · audited 2026-09-17
This skill is a local logging and analysis tool for recording 'improvement insights' and generating weekly reports. The Python code heuristically analyzes conversation length and feedback text, appends supplied insights to a Markdown file, and reads that file to produce reports, statistics, and suggestions. It contains no network requests, no code that accesses credentials or payment methods, and no shell execution. The main caution is local file persistence: it writes to an improvement log under /root/.openclaw/workspace by default, and the package's own risk note says sensitive user content could end up in that log. I found no prompt injection or hidden malicious behavior.
| Declared purpose | Self-improving agent system that analyzes conversation quality, identifies improvement opportunities, and continuously optimizes response strategies. |
| Observed behavior | The class SelfImprovingAgent is instantiated with a workspace path, defaulting to /root/.openclaw/workspace. It defines heuristic analysis based on feedback words and conversation length; logs user-provided improvement insights by appending to improvement_log.md; reads that log for weekly report generation, statistics, and SOUL.md update suggestions. The CLI supports --log, --report, --stats, and --suggest. No network, subprocess, wallet, payment, or credential-access code is present. |
| Verdict, rules only | SAFE |
| Verdict, AI | CAUTION |
| Final verdict | CAUTION (never better than either pass) |
| capability | vs purpose | detail |
|---|---|---|
| filesystem_read | needed | Reads improvement_log.md for weekly reports, stats, and suggestions; also checks whether SOUL.md exists. |
| filesystem_write | needed | Appends timestamped improvement insights and categories to improvement_log.md in the configured workspace. |
| severity | finding |
|---|---|
| low | Hardcoded default workspace path When no workspace is supplied, all log reads/writes default to /root/.openclaw/workspace. A user should confirm this is the intended OpenClaw workspace and that the location is not exposed to unexpected users. self_improving.py:20 def __init__(self, workspace: str = "/root/.openclaw/workspace"): |
| low | Appends to local improvement log log_improvement writes caller-supplied insight text to improvement_log.md. This is required for the skill's declared logging purpose, but the file could accumulate sensitive conversation content if users log secrets. self_improving.py:82 with open(self.improvement_log, "a", encoding="utf-8") as f: |
| info | Reads local improvement log generate_weekly_report reads improvement_log.md to count recent entries. This read is local-only and consistent with the declared weekly reporting feature. self_improving.py:103 with open(self.improvement_log, "r", encoding="utf-8") as f: |
| info | Documentation claims automatic analysis The provided code only analyzes when called through Python or the CLI; no automatic hook, scheduler, or post-session trigger is included. This is a minor documentation discrepancy, not malicious behavior. SKILL.md:33 The skill automatically analyzes conversations after each session. |
| medium | Publisher warns of sensitive local persistence The skill could store sensitive text in the log file if the operator passes such content as insights. This is a known data-handling risk to review before deployment. skill-card.md:25 Risk: Conversation-quality logs and improvement insights may persist private prompts, customer data, credentials, or other sensitive content in the local worksp |
0 AI finding(s) were dropped because their file, line or quote did not match the package.
| domain | service rating | where |
|---|---|---|
| clawhub.ai | not in the service index | skill-card.md:9 |
| github.com | not in the service index | SKILL.md:8 |
| Binaries invoked | bash, python, python3 |
| Environment variables read | none |
| Hard-coded wallets | none |
| Pipes a download to a shell | 0 |
| eval / exec / subprocess | 0 |
| base64 blobs | 0 |
| File writes | 1 |
| Persistence | 0 |
| file | lines | sha256 |
|---|---|---|
SKILL.md | 65 | c4b1d7ccd5c99d2d… |
_meta.json | 6 | caae3e75d3c95ac3… |
self_improving.py | 201 | 5fad6bcc67baf84d… |
skill-card.md | 54 | d3521f2895840061… |
skill.yaml | 10 | 5583ddee06020cb1… |
For agents
JSON: https://agenteconomy.report/k/xiucheng-self-improving-agent.audit.json · badge: https://agenteconomy.report/k/xiucheng-self-improving-agent.audited.svg ·
skill rating: /k/xiucheng-self-improving-agent · 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).