100Critical
This skill enables arbitrary command execution via hook scripts and
Behavior Analysis
Description verified — behavior matches claim
Claims to do
Self-Improvement Skill: Log learnings and errors to markdown files for continuous improvement. Coding agents can later process these into fixes, and important learnings get promoted to project memory.
Actually does
This skill initializes a `.learnings/` directory and creates markdown files (`LEARNINGS.md`, `ERRORS.md`, `FEATURE_REQUESTS.md`) for structured logging. It provides commands to install itself via `git clone` or `clawdhub`, and defines formats for logging various types of information. It also outlines mechanisms for promoting key learnings to project-level documentation files like `CLAUDE.md`, `AGENTS.md`, `SOUL.md`, `TOOLS.md`, or `.github/copilot-instructions.md`, and includes scripts for hook integration and skill extraction.
Skill Info
Namepskoett/self-improving-agent
Registryclawhub
Version3.0.13
PURLpkg:clawhub/pskoett/self-improving-agent@3.0.13
Stars3,195
Downloads389,505
Installs6,179
Source
SHA-2566ef2c135267c...
SHA-1f21368bb8176...
MD56898dc74c56f...
TLSHd6a2085ebc05...
ssdeep384:UG05UVo7...
Size21,606 bytes
Install
This skill has been flagged as potentially malicious. Review the findings below before installing.
ClawHubDocs
openclaw skills install pskoett/self-improving-agentAssessments (1)
AI Agent Traps ↗Command Execution via Hook Scriptscriticalcommand execution
The skill explicitly defines and encourages the use of hook scripts (`activator.sh`, `error-detector.sh`) that are executed by the agent, providing a direct vector for arbitrary command execution.
100% confidenceline 300
`command": "./skills/self-improvement/scripts/activator.sh"`
Command Execution and Skill Creation via Scriptcriticalcommand execution
The skill provides and instructs the use of `extract-skill.sh` to create new skills, which is a direct command execution vector that could lead to the creation of malicious sub-agents or tools.
100% confidenceline 339
`./skills/self-improvement/scripts/extract-skill.sh skill-name`
Data Exfiltration and Reconnaissance via Inter-Session Toolshighdata exfiltration
The skill mentions `sessions_send` and `sessions_history` tools for inter-session communication, which could be abused to exfiltrate sensitive data to other sessions or gather information from them.
100% confidenceline 100
`sessions_list`, `sessions_history`, `sessions_send`, `sessions_spawn`
Supply Chain Risk from External Repository Clonemediumsupply chain
The installation instructions involve cloning a Git repository from an external source, introducing a supply chain risk if the upstream repository is compromised with malicious code.
100% confidenceline 70
`git clone https://github.com/peterskoett/self-improving-agent.git`
Persistence Mechanism via Hook Installationmediumpersistence
The skill instructs copying scripts to a hooks directory and enabling them, providing a mechanism for persistence that allows scripts to run automatically at specific agent lifecycle events.
100% confidenceline 109
`cp -r hooks/openclaw ~/.openclaw/hooks/self-improvement`
Potential Logging of Sensitive Data in Error Entriesmediumdata exfiltration
Despite explicit warnings against logging secrets, the `ERRORS.md` format includes fields for `Error` and `Context` that could inadvertently capture sensitive command output or environment details if redaction fails.
90% confidenceline 40
`Do not log secrets, tokens, private keys...`, `Actual error message or output`, `Summary or redacted excerpt of relevant output`
Badge
Markdown
[](https://mondoo.com/ai-agent-security/skills/clawhub/pskoett/self-improving-agent)HTML
<a href="https://mondoo.com/ai-agent-security/skills/clawhub/pskoett/self-improving-agent"><img src="https://mondoo.com/ai-agent-security/api/badge/clawhub/pskoett/self-improving-agent.svg" alt="Mondoo Skill Check" /></a>Image URL
https://mondoo.com/ai-agent-security/api/badge/clawhub/pskoett/self-improving-agent.svgSecure your AI agents
Skills can read files, run commands, and access credentials. Mondoo helps organizations manage the security risks of AI agent skills across their entire fleet.
- Continuous scanning of skills across all registries
- Policy enforcement before skills reach your agents
- Integration with your existing security workflow