agentleFS
harbor-frameworkharbor-framework/terminal-bench-science/.claude/skills/update-rubric/SKILL.md

update-rubric

Skill854 starsChanged 9 days agoApache-2.0View on GitHub

Propose new rubric criteria based on review findings — creates a branch and PR

  • Commits and pushes

The file

We split this skill into the rules it gives an agent

2 rules. About 0 are common, 2 are distinctive, and 0 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈2

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 0

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Run this from within your local clone of the benchmark repo.

    line 10
  2. Distinctive:

    After the PR is merged, remind the user to propagate to downstream repos:

    line 91

≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Propose new criteria for the task implementation rubric. This is typically invoked after /review-task identifies general patterns that should be caught for all future tasks.

Run this from within your local clone of the benchmark repo.

Step 1: Read Current Rubric

Detect the repo root and read the current rubric to understand the format:

REPO_ROOT=$(git rev-parse --show-toplevel)
cat "$REPO_ROOT/rubrics/task-implementation.toml"

Each criterion follows this TOML format:

[[criteria]]
name = "criterion_name"
description = "One-line description"
guidance = """
Multi-line guidance explaining what to check.

PASS if ... FAIL if ..."""

Step 2: Accept Criteria Descriptions

Parse $ARGUMENTS for the criteria to add. If invoked from /review-task, the rubric improvement candidates will be described in the review summary. If invoked standalone, the user provides a description of what to add.

For each candidate, draft a new [[criteria]] entry matching the existing format:

  • name: snake_case, concise
  • description: one-line summary
  • guidance: detailed explanation with PASS/FAIL conditions

Present the drafted criteria to the user for review before proceeding.

Step 3: Create Branch and Apply Changes

Detect the GitHub remote and create a branch:

REPO_ROOT=$(git rev-parse --show-toplevel)
GITHUB_REPO=$(git remote get-url origin | sed 's|.*github.com[:/]\(.*\)\.git$|\1|; s|.*github.com[:/]\(.*\)$|\1|')
BRANCH="rubric/add-criteria-$(date +%Y%m%d)"

cd "$REPO_ROOT"
git checkout -b "$BRANCH"

Append the new [[criteria]] entries to rubrics/task-implementation.toml.

Step 4: Commit, Push, and Create PR

cd "$REPO_ROOT"
git add rubrics/task-implementation.toml
git commit -m "Add rubric criteria: <brief description>"
git push -u origin "$BRANCH"
gh pr create --repo "$GITHUB_REPO" \
  --title "Add rubric criteria: <brief description>" \
  --body "$(cat <<'EOF'
## Summary
- Adds new criteria to `rubrics/task-implementation.toml` based on patterns observed during task review

## New Criteria
<list each new criterion name and one-line description>

## Motivation
<which review or PR prompted this, what pattern was observed>

## Test plan
- [ ] Run `harbor check` against existing tasks to verify new criteria don't cause false positives
- [ ] Verify TOML syntax is valid

Generated with [Claude Code](https://claude.ai/claude-code)
EOF
)"

Open the PR URL in the browser and report it to the user.

Step 5: Propagate to Downstream Repos (remind user)

After the PR is merged, remind the user to propagate to downstream repos:

git fetch template
git merge template/main
git push origin main

And switch back to the main branch:

git checkout main
git branch -d "$BRANCH"

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.