Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx agentmods add commands/aminry/auto-claude-plugins/evolve-rubricgit clone --depth 1 https://github.com/aminry/auto-claude-pluginsWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/commands/aminry/auto-claude-plugins/evolve-rubric)<a href="https://agentmods.dev/commands/aminry/auto-claude-plugins/evolve-rubric"><img src="https://agentmods.dev/badge/commands/aminry/auto-claude-plugins/evolve-rubric.svg" alt="Measured on agentmods" height="20"></a>What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5 | $0.00014 | $0.02183 |
| Opus 5 | $0.00007 | $0.01092 |
| Sonnet 5 | $0.00003 | $0.00437 |
| Haiku 4.5 | $0.00001 | $0.00218 |
Grade A, and why
evolve-rubric scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 4d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 250 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Analyze the latest auto-improve results to find quality issues not covered by the current evaluation rubric, then update the rubric accordingly.
The auto-improve skill to analyze: $ARGUMENTS
Overview
After running an auto-improve cycle, the evaluation rubric may have blind spots — quality issues that exist in the output but aren't captured by any rubric question. This skill:
- Reads the latest extraction results from the most recent auto-improve run
- Performs a deep, rubric-independent quality analysis
- Identifies gaps where real issues exist but no rubric question catches them
- Proposes new rubric questions and gets user approval
- Updates the evaluation rubric
Phase 1: Load Context
-
Identify the feature: Parse
$ARGUMENTSto determine the auto-improve skill name. Expected format: a feature name likecomparisons,products,features, etc. -
Load the current rubric: Read
.claude/commands/evaluate-{feature}.md(or the appropriate evaluate file — check.claude/commands/for the matching evaluate skill). -
Load the latest results: Find the debug/output directory by reading the auto-improve skill file (
.claude/commands/auto-improve-{feature}.md) and locating its configured paths (typically.debug/auto-improve-{feature}/but may vary). Read the most recent extraction files from the extractions subdirectory. Sort by modification time and take the latest iteration's files. -
Load the results TSV: Read the
results.tsvfile from the auto-improve output directory to understand the score trajectory and which failures were already addressed. -
Load the latest evaluations: Read the evaluations subdirectory for the most recent iteration to see which rubric questions passed/failed.
If any of these files don't exist, report what's missing and stop.
Phase 2: Deep Quality Analysis
For each extraction file from the latest iteration, perform a rubric-independent quality review. This means: forget the rubric exists and evaluate the output fresh, looking for ANY quality issue regardless of whether a rubric question covers it.
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 4d ago First seen · 250 lines · 14 tokens per session scan A 311a71eacaa6
evolve-rubric is a command published in the GitHub repository aminry/auto-claude-plugins (1 stars, last pushed 3mo ago), licensed MIT. It adds 14 tokens to every session and 2,183 once invoked, about $0.0001 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other commands, from other repositories
perf
Structured performance investigation with baselines, profiling, and evidence-backed decisions.
audit-project-agents
This file contains detailed agent coordination for /audit-project.
ship-ci-review-loop
Command "ship-ci-review-loop" from composio-community/awesome-claude-plugins, covering phase 4: ci & review monitor loop - reference, this phase is mandatory, pr auto-review process, why all comments matter and the monitor loop algorithm.
ship
Complete PR workflow from commit to production with validation.
audit-project
Multi-agent code review with iterative improvement.
help
Show skill-bus overview, status, and available commands. Quick reference for all skill-bus features.