Claude Skill

jevaluate-private

Rates a project with Jevaluate and keeps the rating private, with the same rubric and checks, a private library that export refuses, and nothing pushed or published. Use when a rating must stay private, such as "/jevaluate-private" followed by a link or folder, "rate this private

LLM Mart · 0 points · 1 views 0 listing impressions 0 install-command copies
Download tiffygk-jev-mode-jevaluate-private-30a0a86.zip · 2 KB

Install

skills CLI npx skills add https://github.com/tiffygk/jev-mode/tree/main/jevaluate-private
Claude Code claude plugin marketplace add https://llmmart.ai/marketplace.json && claude plugin install tiffygk-jev-mode@llmmart
Git git clone https://github.com/tiffygk/jev-mode.git

The skills CLI installs just this skill, for any of its supported agents. Claude Code installs the whole tiffygk/jev-mode collection as a plugin from our marketplace. Git is the plain clone.

README

Jevaluate Private

Updated private version of Jevaluate for how devs and teams build and eval their own projects. Rate a project you can't publish: a client's codebase, your employer's pipeline, a design you were sent, or your own work in progress. It uses the same rubric, steps and checks as Jevaluate Publish, so a private verdict means the same as a public one. The rating stays on your machine, so the skill skips the publishing steps.

After receiving your Jevaluate Private rating:

  • Read the verdict and tune the rating to your own product judgment by talking it through with the skill. The rater may have missed some of your API calls, or made calls you don't agree with.
  • Run Jevaluate Report to write a Markdown file of the proposed fixes, which your agents build and test in your Jev project.

How is this different from Jevaluate Publish?

Jevaluate Publish rates a project in full and publishes the rating to this repo's ratings page, so the community can find the quality Jev integrations. Its ratings shouldn't be tuned, since they are an unbiased eval for the community.

Use it

Ask your agent to run jevaluate-private https://github.com/you/project, or give it a folder: jevaluate-private ~/work/retrieval-graph. "Rate this privately" works too.

How it stays private

  • Private ratings live in their own library (~/.jevaluate-private/ by default), never the public one.
  • The skill refuses to log a private rating into a public library, or to export a private library or rating.
  • A local folder is read from disk. Nothing in it is fetched from or sent to GitHub.

Install

Needs Claude Code or Codex, Jevaluate Publish, Python 3.8+, and the GitHub CLI only when you rate a GitHub link. In Codex, approve the prompt to save to the private library, which sits outside your project folder.

In Claude Code:

claude plugin marketplace add tiffygk/jev-mode
claude plugin install jevaluate-private@jev-mode
claude plugin install jevaluate@jev-mode

In Codex:

codex plugin marketplace add tiffygk/jev-mode
codex plugin add jevaluate-private@jev-mode
codex plugin add jevaluate@jev-mode

To install every jev-mode skill at once, or to copy skills by hand, see the repo README.

Skill manifest

Jevaluate Private

Rates a project exactly as jevaluate-publish does, with the same rubric, steps and checks, and keeps the rating out of everything public. The rating lives in a private library that export refuses, and nothing is pushed, published or added to ratings/.

Setup, once per machine

python3 <jevaluate>/scripts/library.py init-private ~/.jevaluate-private creates the private library, marked by a PRIVATE file. <jevaluate> is the jevaluate/ folder, beside this one.

How to rate

  1. Set JEVALUATE_LIBRARY=~/.jevaluate-private for every command in the rating, so step.py and library.py read and write only the private library. Create the library first if it doesn't exist.
  2. Gather the files:
    • a GitHub link: coverage_manifest.py owner/repo <dir>, as usual;
    • a folder on disk: coverage_manifest.py --local <folder> <dir>. It applies the same skip rules and size cap, reads nothing outside the folder, and records git HEAD as the commit (or a content fingerprint when the folder isn't a clean git checkout).
  3. Follow <jevaluate>/SKILL.md and its read.md unchanged. Two differences only:
    • the project's own files may be private, while the rules still come only from docs.typesafe.ai;
    • a local folder's url is local:<short-name>.
  4. Add visibility: private to the rating's front matter, then log it with ../jevaluate/scripts/library.py add as usual. add refuses a private rating into a public library, and refuses a non-private rating into a private one.
  5. Finish with the rating's verdict and its fixes in plain words, for the person who asked. Offer a page if they want one to share. If they agree with the rating, offer jevaluate-report: it writes a file of the fixes, each with a test, for their coding agents to work through before a pull request, saved where they choose. Never run ../jevaluate/scripts/library.py export, never copy the rating into a repo, and never push.

Comparisons with past ratings come only from the private library, so a private rating is never shown beside a public one.

In Codex, follow the Codex note in <jevaluate>/SKILL.md. The private library sits outside the workspace, so library.py add and init-private need the user's approval. Ask for it; never move the library to a temporary folder to avoid the prompt.

Enforced by

Rule Enforced by
A private rating goes only into a private library, and a private library takes only private ratings ../jevaluate/scripts/library.py add; ../jevaluate/scripts/test_private.py
A rating of a local folder counts as private ../jevaluate/scripts/library.py add; ../jevaluate/scripts/test_private.py
Export refuses a private library and any private rating ../jevaluate/scripts/library.py export; ../jevaluate/scripts/test_private.py
A local folder is read with the same skip rules, and nothing outside it ../jevaluate/scripts/coverage_manifest.py --local; ../jevaluate/scripts/test_private.py
JEVALUATE_LIBRARY set for every command; no rating copied into a repo or pushed judgment
Files (jev-mode)
  • README.md 2.3 KB
    # Jevaluate Private
    
    Updated private version of Jevaluate for how devs and teams build and eval their own projects. Rate a project you can't publish: a client's codebase, your employer's pipeline, a design you were sent, or your own work in progress. It uses the same rubric, steps and checks as [Jevaluate Publish](../jevaluate/), so a private verdict means the same as a public one. The rating stays on your machine, so the skill skips the publishing steps.
    
    After receiving your Jevaluate Private rating:
    - Read the verdict and tune the rating to your own product judgment by talking it through with the skill. The rater may have missed some of your API calls, or made calls you don't agree with.
    - Run [Jevaluate Report](../jevaluate-report/) to write a Markdown file of the proposed fixes, which your agents build and test in your Jev project.
    
    **How is this different from Jevaluate Publish?**
    
    Jevaluate Publish rates a project in full and publishes the rating to this repo's ratings page, so the community can find the quality Jev integrations. Its ratings shouldn't be tuned, since they are an unbiased eval for the community.
    
    ## Use it
    
    Ask your agent to run `jevaluate-private https://github.com/you/project`, or give it a folder: `jevaluate-private ~/work/retrieval-graph`. "Rate this privately" works too.
    
    ## How it stays private
    
    - Private ratings live in their own library (`~/.jevaluate-private/` by default), never the public one.
    - The skill refuses to log a private rating into a public library, or to export a private library or rating.
    - A local folder is read from disk. Nothing in it is fetched from or sent to GitHub.
    
    ## Install
    
    Needs Claude Code or Codex, [Jevaluate Publish](../jevaluate/), Python 3.8+, and the GitHub CLI only when you rate a GitHub link. In Codex, approve the prompt to save to the private library, which sits outside your project folder.
    
    In Claude Code:
    
    ```
    claude plugin marketplace add tiffygk/jev-mode
    claude plugin install jevaluate-private@jev-mode
    claude plugin install jevaluate@jev-mode
    ```
    
    In Codex:
    
    ```
    codex plugin marketplace add tiffygk/jev-mode
    codex plugin add jevaluate-private@jev-mode
    codex plugin add jevaluate@jev-mode
    ```
    
    To install every jev-mode skill at once, or to copy skills by hand, see the [repo README](../README.md#install).
    
  • SKILL.md 3.7 KB
    ---
    name: jevaluate-private
    description: Rates a Jev project and keeps the rating private, with the same rubric and checks as Jevaluate, a verdict and fixes, a private library that export refuses, and nothing pushed or published. It is the default way to rate a project being built. Use when someone wants their own, their team's or a client's project rated, such as "rate my project", "jevaluate my repo", "rate our Jev integration", "/jevaluate-private" followed by a link or folder, "rate this privately", "evaluate my own project", a client's or employer's code, a design doc or a local folder that isn't public, or any rating that must not reach the public ratings or the repo.
    ---
    
    # Jevaluate Private
    
    Rates a project exactly as `jevaluate-publish` does, with the same rubric, steps and checks, and keeps the rating out of everything public. The rating lives in a private library that export refuses, and nothing is pushed, published or added to `ratings/`.
    
    ## Setup, once per machine
    
    `python3 <jevaluate>/scripts/library.py init-private ~/.jevaluate-private` creates the private library, marked by a `PRIVATE` file. `<jevaluate>` is the `jevaluate/` folder, beside this one.
    
    ## How to rate
    
    1. Set `JEVALUATE_LIBRARY=~/.jevaluate-private` for every command in the rating, so `step.py` and `library.py` read and write only the private library. Create the library first if it doesn't exist.
    2. Gather the files:
       - a GitHub link: `coverage_manifest.py owner/repo <dir>`, as usual;
       - a folder on disk: `coverage_manifest.py --local <folder> <dir>`. It applies the same skip rules and size cap, reads nothing outside the folder, and records git HEAD as the commit (or a content fingerprint when the folder isn't a clean git checkout).
    3. Follow `<jevaluate>/SKILL.md` and its `read.md` unchanged. Two differences only:
       - the project's own files may be private, while the rules still come only from docs.typesafe.ai;
       - a local folder's `url` is `local:<short-name>`.
    4. Add `visibility: private` to the rating's front matter, then log it with `../jevaluate/scripts/library.py` add as usual. `add` refuses a private rating into a public library, and refuses a non-private rating into a private one.
    5. Finish with the rating's verdict and its fixes in plain words, for the person who asked. Offer a page if they want one to share. If they agree with the rating, offer `jevaluate-report`: it writes a file of the fixes, each with a test, for their coding agents to work through before a pull request, saved where they choose. Never run `../jevaluate/scripts/library.py` export, never copy the rating into a repo, and never push.
    
    Comparisons with past ratings come only from the private library, so a private rating is never shown beside a public one.
    
    In Codex, follow the Codex note in `<jevaluate>/SKILL.md`. The private library sits outside the workspace, so `library.py add` and `init-private` need the user's approval. Ask for it; never move the library to a temporary folder to avoid the prompt.
    
    ## Enforced by
    
    | Rule | Enforced by |
    |---|---|
    | A private rating goes only into a private library, and a private library takes only private ratings | `../jevaluate/scripts/library.py` add; `../jevaluate/scripts/test_private.py` |
    | A rating of a local folder counts as private | `../jevaluate/scripts/library.py` add; `../jevaluate/scripts/test_private.py` |
    | Export refuses a private library and any private rating | `../jevaluate/scripts/library.py` export; `../jevaluate/scripts/test_private.py` |
    | A local folder is read with the same skip rules, and nothing outside it | `../jevaluate/scripts/coverage_manifest.py` --local; `../jevaluate/scripts/test_private.py` |
    | `JEVALUATE_LIBRARY` set for every command; no rating copied into a repo or pushed | judgment |
    

Comments (0)

Sign in to join the conversation.

No comments yet.

Reviews (0)

No reviews yet.