← Back to Skills Marketplace
131
Downloads
0
Stars
0
Active Installs
1
Versions
Install in OpenClaw
/install research-loop
Description
Claude Code compatibility mirror for the Codex-native 10000 Mentors Research Workflow. Use only when running from Claude Code and the user wants the same sou...
Usage Guidance
This skill appears coherent for running a repo-focused research loop. Before installing or running it: (1) confirm the runtime has python3, git, and the GitHub CLI installed and authenticated if you expect publishing; GH/git will use whatever credentials are already configured, so be mindful that pushes or gh actions could occur under those identities. (2) Verify the `autonomous_research_workflow` CLI/module the skill invokes is present in the runtime or in the target repo to avoid the agent attempting to download or install code at runtime. (3) Because the skill writes changes into a `source_changes/` mirror and may publish via GitHub, consider trying it first in a sandbox or on a fork to observe behavior. (4) If you do not want any remote publishing, ensure the environment has no active gh auth or run with a local-only dry-run. Overall the skill is internally consistent, but watch for implicit use of existing git/gh credentials and ensure you trust the runtime environment.
Capability Analysis
Type: OpenClaw Skill
Name: research-loop
Version: 1.0.0
The research-loop skill bundle defines a highly structured, 15-phase autonomous research workflow focused on scientific rigor, reproducibility, and 'protocol hygiene.' While it requests broad permissions including Bash, WebSearch, and the GitHub CLI (gh), these tools are consistent with its stated purpose of automating research, running experiments, and publishing results. The instructions (SKILL.md and references/) are designed to prevent common AI agent failures, such as making misleading claims or getting stuck in 'asset-polishing' loops, rather than to exfiltrate data or establish persistence. No evidence of malicious intent, obfuscation, or unauthorized data access was found.
Capability Tags
Capability Assessment
Purpose & Capability
Name/description describe a repo-focused research loop; declared required binaries (python3, git, gh) and allowed tools (Bash, WebSearch, Read/Write) are consistent with reading a repository, running Python helpers, and publishing to GitHub. There are no unrelated credentials or binaries requested.
Instruction Scope
SKILL.md instructs the agent to read the target repo, run hygiene and frontier checks, produce a single bounded micro-step, write changes into source_changes/, and emit an executor_manifest via `python3 -m autonomous_research_workflow.cli`. All referenced files and paths are within the workflow's repository domain. The instructions explicitly forbid using unrelated APIs (e.g., Ollama). They do not request arbitrary host files or environment secrets beyond normal repo operations.
Install Mechanism
This is an instruction-only skill with no install spec or archive downloads. No packages are fetched or written by the skill itself, minimizing disk-write/install risk. The only implicit requirement is that the runtime provide the declared binaries and the Python module/CLI referenced by the instructions (expected to be in the runtime or repo).
Credentials
The skill declares no required env vars or credentials, which is appropriate. One operational note: the GitHub CLI ('gh') and git operate using whatever credentials are already configured in the runtime/user environment; the skill does not explicitly request tokens but can implicitly act using existing auth. That behavior is expected for a GitHub-publishing workflow but is worth awareness.
Persistence & Privilege
The skill is not always-enabled and is user-invocable. It does not request permanent platform privileges or attempt to modify other skills or global agent settings. Autonomous invocation (model calls) remains allowed (platform default) but is not combined with elevated 'always' privilege or unexplained credential access.
How to Use
- Make sure OpenClaw is installed (local or Docker)
- Run the install command in chat:
/install research-loop - After installation, invoke the skill by name or use
/research-loop - Provide required inputs per the skill's parameter spec and get structured output
Version History
v1.0.0
- Initial release of the research-loop skill, providing a Claude Code compatibility mirror for the 10000 Mentors Research Workflow.
- Maintains the same source-gated research loop contract as the Codex-native version, tailored specifically for the Claude Code runtime.
- Enforces strict protocol hygiene, innovation frontier checks, and phase order as described in the upstream reference documentation.
- Requires only Python 3, Git, and GitHub CLI (gh); no dependence on Ollama APIs or services.
- Guides each research loop through a 15-phase process, from source intake to GitHub publishing, ensuring rigorous standards before completion.
Metadata
Frequently Asked Questions
What is Research Loop?
Claude Code compatibility mirror for the Codex-native 10000 Mentors Research Workflow. Use only when running from Claude Code and the user wants the same sou... It is an AI Agent Skill for Claude Code / OpenClaw, with 131 downloads so far.
How do I install Research Loop?
Run "/install research-loop" in the OpenClaw or Claude Code chat to install it in one step — no extra setup required.
Is Research Loop free?
Yes, Research Loop is completely free, licensed under MIT-0. You can download, install and use it at no cost.
Which platforms does Research Loop support?
Research Loop is cross-platform and runs anywhere OpenClaw / Claude Code is available (cross-platform).
Who created Research Loop?
It is built and maintained by Da Wei (@wd041216-bit); the current version is v1.0.0.
More Skills