llm-as-judge
Agent skill that has an independent model score the last turn from 0 to 10.
Safety scan
A static text scan, not a guarantee. Select the badge to see what was scanned and every finding.
Badge for your README
Markdown
[](/listings/orenatobr--llm-as-judge)HTML
<a href="/listings/orenatobr--llm-as-judge"><img src="/api/listings/orenatobr--llm-as-judge/badge.svg" alt="findsafeskills safety rating" /></a>Install
Install instructions from the README
As written by the repo's author. findsafeskills has not verified these steps.
For Claude Code, copy the skill folders into your personal skills directory:
git clone https://github.com/orenatobr/llm-as-judge.git mkdir -p ~/.claude/skills cp -R llm-as-judge/skills/* ~/.claude/skills/ mkdir -p ~/.claude/scripts cp llm-as-judge/scripts/* ~/.claude/scripts/ chmod +x ~/.claude/scripts/*.sh
Start a new Claude Code session. Skills load from the `description` in each `SKILL.md`, so they trigger on their own and can also be called by name.
Other agents: each skill is a folder with a `SKILL.md` (YAML frontmatter plus Markdown). Tools that read the Agent Skills format can load `skills/` as it is. For the rest, paste the body of the `SKILL.md` into the agent's instruction file.
What was scanned
The files the safety scan read, as written by the repository's author and shown as plain text. Lines with a finding are marked.
README.md · README · 49 lines
- # llm-as-judge
- An independent model scores the agent's last turn on what code cannot check: whether the answer follows the request, whether the tone fits, and whether a diff does what the request asked. Below 7 the agent iterates on the feedback. At 7 or above it moves on. `scripts/llm-judge.sh` calls the `claude` CLI with Claude Fable 5.1 and falls back to Opus 5.5. It runs only after `quality-gate` passed for the same turn.
- ## Install
- For Claude Code, copy the skill folders into your personal skills directory:
- ```bash
- git clone https://github.com/orenatobr/llm-as-judge.git
- mkdir -p ~/.claude/skills
- cp -R llm-as-judge/skills/* ~/.claude/skills/
- mkdir -p ~/.claude/scripts
- cp llm-as-judge/scripts/* ~/.claude/scripts/
- chmod +x ~/.claude/scripts/*.sh
- ```
- Start a new Claude Code session. Skills load from the `description` in each `SKILL.md`, so they trigger on their own and can also be called by name.
- Other agents: each skill is a folder with a `SKILL.md` (YAML frontmatter plus Markdown). Tools that read the Agent Skills format can load `skills/` as it is. For the rest, paste the body of the `SKILL.md` into the agent's instruction file.
- ## Usage
- Run `/quality-gate` first, then `/llm-as-judge`. The skill passes `--state`, `--turn` and `--transcript` to the script. A Stop hook can automate both, but this repo does not ship one.
- ## Requirements
- `claude` CLI on PATH, `jq`, `git` and `python3`. Install [quality-gate](https://github.com/orenatobr/quality-gate) as well, because the judge refuses to run until the gate has passed.
- ## Related skills
- [quality-gate](https://github.com/orenatobr/quality-gate) (must run first).
- ## What is in this repo
- | Path | Purpose |
- |------|---------|
- | `skills/llm-as-judge/` | The `llm-as-judge` skill. |
- | `scripts/` | Helper scripts that the skill calls. |
- | `CHANGELOG.md` | Notable changes, in Keep a Changelog format. |
- | `AGENTS.md` | Rules for AI agents that edit this repo. |
- ## Contributing
- `main` is protected: changes land through a pull request with one approval. See [AGENTS.md](AGENTS.md) for the conventions.
- ## License
- MIT, see [LICENSE](LICENSE).
Ratings and comments
Sign in to rate this skill.
Sign inFree to use. If it’s helped, pass it on.