← Back to Skills Marketplace
142
Downloads
0
Stars
0
Active Installs
1
Versions
Install in OpenClaw
/install perrytest
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 be what it says: a local helper that logs errors/learnings and can inject brief reminders into OpenClaw sessions. Before enabling or installing: 1) Review the scripts (activator.sh, error-detector.sh, extract-skill.sh) yourself — they run locally and create files in your workspace; 2) Note the error detector reads CLAUDE_TOOL_OUTPUT (tool outputs can include secrets), so enable PostToolUse hooks only if you trust the runtime and want reminders; 3) The hook files copy into ~/.openclaw/hooks and, if enabled, will inject a small virtual file at bootstrap — this changes session context (intentionally). If you’re unsure, install manually or enable hooks only at project-level (not global) and verify file permissions; finally, if you plan to trust the upstream GitHub repo, review its history and author before cloning.
Capability Analysis
Type: OpenClaw Skill
Name: perrytest
Version: 1.0.0
The skill bundle implements a self-improvement framework that allows an AI agent to log errors, corrections, and new knowledge into local markdown files (e.g., `.learnings/LEARNINGS.md`). It includes shell scripts (`activator.sh`, `error-detector.sh`) and OpenClaw hooks (`handler.js`) designed to inject reminders into the agent's context and detect command failures. The `extract-skill.sh` script provides a utility for scaffolding new skills from captured learnings with basic path sanitization to prevent directory traversal. All behaviors are transparent, well-documented, and strictly aligned with the stated purpose of continuous agent improvement without any evidence of malicious intent or data exfiltration.
Capability Assessment
Purpose & Capability
Name/description (capture learnings/errors and promote them into workspace files) lines up with the provided scripts and hook handlers. The repo files implement reminders, an error detector that watches tool output, and a scaffold script to extract skills from learnings — all coherent with the stated purpose.
Instruction Scope
Runtime instructions focus on creating/using a .learnings/ folder and optionally copying/enabling the OpenClaw hook. Scripts: activator.sh prints a short reminder; error-detector.sh reads CLAUDE_TOOL_OUTPUT and emits a reminder when error patterns match; handler.js/ts inject a virtual file into bootstrapFiles. The only environment variable referenced at runtime is CLAUDE_TOOL_OUTPUT (expected for PostToolUse hooks) which is not declared in requires.env but is an agent-provided context variable. No instructions attempt to read unrelated system files, exfiltrate data, or call external endpoints.
Install Mechanism
No automated install spec in the registry; SKILL.md suggests manual git clone or ClawdHub install and optionally copying hook files into ~/.openclaw. The recommended sources are GitHub repos (public), and the included scripts are small local shell/JS files. Nothing downloads arbitrary archives or runs remote code automatically during install.
Credentials
The skill declares no required env vars or credentials. The error-detector.sh uses the CLAUDE_TOOL_OUTPUT environment variable (agent-provided) to detect failures — reasonable for its function but worth noting since it inspects tool output (which can contain sensitive command output). The extract-skill.sh writes scaffolds into the current workspace (it enforces no absolute/.. paths) — write access is intentional and proportional to creating local skills.
Persistence & Privilege
always:false and no requests to modify other skills or global configs. Hooks are optional and only run if the user copies/enables them. The hook/handler injects a small virtual bootstrap file (prompt-reminder) — this is prompt injection by design but limited in scope and local to the workspace/hook lifecycle.
How to Use
- Make sure OpenClaw is installed (local or Docker)
- Run the install command in chat:
/install perrytest - After installation, invoke the skill by name or use
/perrytest - Provide required inputs per the skill's parameter spec and get structured output
Version History
v1.0.0
Skill initial release: enables structured self-improvement logging for agent workspaces.
- Introduces detailed guidelines for logging errors, corrections, knowledge gaps, feature requests, and best practices to dedicated markdown files under `.learnings/`.
- Outlines workflows to promote broadly applicable learnings to permanent workspace memory (e.g., `AGENTS.md`, `SOUL.md`, `TOOLS.md`, `CLAUDE.md`).
- Provides ready-to-use templates for Learning, Error, and Feature Request entries, including metadata structure and status management.
- Documents integration for OpenClaw and generic agent setups, with setup steps and prompt/hook recommendations.
- Describes inter-session learning sharing utilities and instructions for automating reminders on session start.
Metadata
Frequently Asked Questions
What is perrytest?
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 142 downloads so far.
How do I install perrytest?
Run "/install perrytest" in the OpenClaw or Claude Code chat to install it in one step — no extra setup required.
Is perrytest free?
Yes, perrytest is completely free, licensed under MIT-0. You can download, install and use it at no cost.
Which platforms does perrytest support?
perrytest is cross-platform and runs anywhere OpenClaw / Claude Code is available (cross-platform).
Who created perrytest?
It is built and maintained by Lipairui (@lipairui); the current version is v1.0.0.
More Skills