← Back to Skills Marketplace
Self Improvement Tianjin
by
tianjin-ren
· GitHub ↗
· v1.0.1
· MIT-0
198
Downloads
0
Stars
1
Active Installs
2
Versions
Install in OpenClaw
/install self-improvement-tianjin
Description
Captures learnings, errors, and corrections to enable continuous improvement. Use when: (1) A command or operation fails unexpectedly, (2) User corrects Clau...
Usage Guidance
This skill appears to do what it claims: collect and promote learnings via local files and optional hooks. Before enabling hooks or running scripts, review the shipped scripts (activator.sh, error-detector.sh, extract-skill.sh, handler.js/ts) to confirm you’re comfortable with them running in your environment. Pay attention to: (1) Naming inconsistencies in docs/paths (e.g., 'self-improving-agent' vs 'self-improvement-tianjin') — ensure you copy to the intended directory; (2) Hooks run with the agent's filesystem permissions and error-detector reads CLAUDE_TOOL_OUTPUT (which can contain command output that may be sensitive) — only enable if you’re okay with that local data being processed; (3) extract-skill.sh will create files under the current workspace (it contains checks to avoid absolute/parent dirs — still review intended output dir). If you plan to install from the referenced GitHub repo, verify the repository URL and contents match the packaged files. If you prefer minimal risk, use the manual workflow (create .learnings/ and copy templates) without enabling the automatic hooks.
Capability Analysis
Type: OpenClaw Skill
Name: self-improvement-tianjin
Version: 1.0.1
The bundle implements a 'self-improvement' workflow that allows AI agents to log errors, corrections, and learnings to local markdown files within a `.learnings/` directory. The included bash scripts (scripts/activator.sh, scripts/error-detector.sh, and scripts/extract-skill.sh) are standard utilities for context injection and file scaffolding; notably, extract-skill.sh includes basic security checks to prevent path traversal. The OpenClaw hooks (hooks/openclaw/handler.js/ts) are used legitimately to inject reminders into the agent's bootstrap process, and no evidence of data exfiltration, malicious execution, or harmful prompt injection was found.
Capability Assessment
Purpose & Capability
The skill's name/description (capture learnings, errors, promote to workspace) match the included artifacts: logging templates, hook handler, activator/error-detector scripts, and an extraction helper. The presence of hook scripts and a skill-extraction script is expected for this purpose. Minor naming inconsistencies exist (registry slug/version vs SKILL.md name vs example install paths: 'self-improvement-tianjin' / 'self-improvement' / 'self-improving-agent'), which are likely bookkeeping inconsistencies but not functionally harmful.
Instruction Scope
SKILL.md instructs creating .learnings files, copying/enabling an OpenClaw hook, and optionally registering scripts as hooks. The scripts only read local tool output (error-detector reads CLAUDE_TOOL_OUTPUT) and append virtual bootstrap files via the hook handler; they do not transmit data externally. The instructions do reference workspace paths (e.g., ~/.openclaw) and recommend copying files to those locations — expected for integration but requires user action/consent.
Install Mechanism
There is no automatic install spec; installation is manual (git clone or clawdhub). All shipped scripts are plain shell/TS/JS files and use local filesystem operations. No remote downloads or opaque installers are used. This is low-to-moderate risk but appropriate for a local integration skill.
Credentials
The skill declares no required env vars or credentials. The error-detector script reads CLAUDE_TOOL_OUTPUT (an agent-provided runtime variable) — that is appropriate for detecting command failures. No unrelated secrets or cloud credentials are requested.
Persistence & Privilege
always:false and disable-model-invocation:false (normal). Hooks are optional and must be enabled by the user; enabling them will cause the scripts to run with the same permissions as the agent process. This is expected for hooks, but users should be aware that hook scripts run locally with agent permissions.
How to Use
- Make sure OpenClaw is installed (local or Docker)
- Run the install command in chat:
/install self-improvement-tianjin - After installation, invoke the skill by name or use
/self-improvement-tianjin - Provide required inputs per the skill's parameter spec and get structured output
Version History
v1.0.1
- Metadata updated in _meta.json; no changes made to functionality or documentation.
- Version incremented to 1.0.1.
v1.0.0
Initial release of the self-improvement skill for OpenClaw and generic coding agents.
- Provides a structured workflow for logging errors, learnings, and feature requests to markdown files for continuous improvement.
- Outlines clear logging formats and promotion targets for broadly applicable learnings (e.g., SOUL.md, AGENTS.md, TOOLS.md).
- Includes detailed setup instructions for both OpenClaw and agent-agnostic environments.
- Introduces standardized log entry formats with ID generation and resolution tracking.
- Supports inter-session communication and hook-based reminders within OpenClaw.
Metadata
Frequently Asked Questions
What is Self Improvement Tianjin?
Captures learnings, errors, and corrections to enable continuous improvement. Use when: (1) A command or operation fails unexpectedly, (2) User corrects Clau... It is an AI Agent Skill for Claude Code / OpenClaw, with 198 downloads so far.
How do I install Self Improvement Tianjin?
Run "/install self-improvement-tianjin" in the OpenClaw or Claude Code chat to install it in one step — no extra setup required.
Is Self Improvement Tianjin free?
Yes, Self Improvement Tianjin is completely free, licensed under MIT-0. You can download, install and use it at no cost.
Which platforms does Self Improvement Tianjin support?
Self Improvement Tianjin is cross-platform and runs anywhere OpenClaw / Claude Code is available (cross-platform).
Who created Self Improvement Tianjin?
It is built and maintained by tianjin-ren (@tianjin-ren); the current version is v1.0.1.
More Skills