analyze

A study-results review guide for checking whether an experiment’s comparisons are fair and choosing whether to adopt a configuration.

In plain words
What is it for?
Use it to review parameter-sweep results, inspect search coverage and training curves, identify invalid or infeasible runs, and decide whether a change or configuration should be adopted.
Why use it?
It prevents conclusions based on missing data, unfair comparisons, overfitting, or unstable results. It also states when the available evidence is insufficient.

Skill for Claude CodeCodex

Install

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.

agentmods
npx agentmods add skills/emaballarin/ccplugins/analyze
Any agent
npx skills add emaballarin/ccplugins --skill analyze
Clone the repo
git clone --depth 1 https://github.com/emaballarin/ccplugins

Made for: Claude Code, Codex.

Per session 176 Skills are progressive disclosure: only the name and description are preloaded; the body loads when the skill is used.
When invoked 1,439 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. Scan, not verified.
Origin original No closer match found in the catalogue.
Token cost

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.

ModelPer sessionOnce invoked
Fable 5 $0.00176 $0.01439
Opus 5 $0.00088 $0.00720
Sonnet 5 $0.00035 $0.00288
Haiku 4.5 $0.00018 $0.00144

Measured yesterday against content hash 61edd958e7b1, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

analyze 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 yesterday.

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.

plugins/tuneml/skills/analyze/SKILL.md · 121 lines

How it starts

The opening of the file, as written. The whole thing — 121 lines — stays where its author put it; the contents beside it link to each section on GitHub.

/tml:analyze — what the study actually showed

Run the checklist before answering the round's question, because several of its entries can invalidate the round, and nobody wants to hear that after being told the answer.

First action, always

ls -la ./.tml/rounds/ 2>/dev/null | tail -5

Then establish which mode you are in:

  • Designed — a study spec exists in ./.tml/rounds/NNN/. Read it. The role assignment and the fixed-hyperparameter caveats are what make the fairness question answerable.
  • Standalone — results only, no spec. This is a normal mode, not a degraded one (references/regime.md §4). Ask which hyperparameters the question is about; treat the rest as unknown-role; and answer the fairness question "cannot be determined" rather than "yes".

Hard rules

  1. Checklist before conclusion. In order, and reported even when it passes.
  2. Never invent a role assignment you were not given. "Cannot be determined" is a real finding — it means the comparison's fairness is unverified.
  3. Name the disabled checks. Missing curves disable overfitting and late-variance detection; a missing infeasibility flag disables §4. Silence about a check that could not run reads as a check that passed.
  4. Read-first. Writes only under ./.tml/. Never edits project code.

Procedure

1. Ingest

The expected shape is templates/results-example.jsonl. Coerce a CSV or tracker export into it. Required per trial: an identifier, the hyperparameters, and the objective. Optional, and each one gates a check: the metric-vs-step series, the best-step, the infeasibility flag and reason, the seed, wall-clock.

Say what was ingested and what was absent before analysing anything.

2. The checklist — references/diagnostics.md §1

  1. Search-space boundaries (§2) — plot the objective against each varied hyperparameter. Best points hugging a bound means the space decided the answer; expand and re-run. If everything above some learning rate is infeasible and the best trials sit at that edge, stop and go to references/instability.md — that is a stability defect wearing an optimum's clothes.
  2. Sampling density (§3) — no general answer exists; say so, and show how many points landed in the good region.
  3. Infeasible fraction (§4) — a large fraction means a bad space or a bug. Report it as a number with reasons, never as missing rows.
  4. Optimisation failuresreferences/instability.md.
  5. Training curves (§5) — problematic overfitting, late step-to-step variance, still-improving, saturated-early, or training loss rising (a bug). Check the best trial of every scientific setting, not just the overall best, and look at the whole population: selecting the winner suppresses overfitting and quietly rewards configurations that were merely hobbled.
  6. Was the nuisance tuning good enough to make the comparison fair (references/study-design.md §4)?

Read the full file on GitHub · 121 lines

Changes

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.

  1. yesterday First seen · 121 lines · 176 tokens per session scan A 61edd958e7b1

Subscribe to this mod's changes

analyze is a skill published in the GitHub repository emaballarin/ccplugins (3 stars, last pushed 26d ago), licensed MIT. It adds 176 tokens to every session and 1,439 once invoked, about $0.0009 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.

Related

Other skills, from other repositories

systematic-debugging

Use when encountering any bug, test failure, or unexpected behavior, before proposing fixes.

obra/superpowers · 21 tokens

next-cache-components-adoption

Turn on Cache Components in a Next.js app and resolve the blocking routes it surfaces. Use when the user wants to enable, adopt, or migrate to Cache Components, flip the cacheComponents flag, work through a flood of blocking-prerender / instant validation errors, run the cache-components-instant-false codemod, or…

vercel/next.js · 95 tokens

babysit-pr

Babysit a GitHub pull request after creation by continuously polling review comments, CI checks/workflow runs, and mergeability state until the PR is merged/closed or user help is required. Diagnose failures, retry likely flaky failures up to 3 times, auto-fix/push branch-related issues when appropriate, and keep…

openai/codex · 114 tokens

imagegen

Generate or edit raster images when the task benefits from AI-created bitmap visuals such as photos, illustrations, textures, sprites, mockups, or transparent-background cutouts. Use when Codex should create a brand-new image, transform an existing image, or derive visual variants from references, and the output…

openai/codex · 113 tokens

cpu-profile-analysis

Analyze V8/Chrome CPU profiles (.cpuprofile) and DevTools trace files (Trace-.json). Use when: profiling performance, investigating slow functions, comparing code paths, finding bottlenecks, analyzing timeToRequest, understanding call trees from sampling profiler data, analyzing layout/paint/rendering, investigating…

microsoft/vscode · 71 tokens

next-cache-components-optimizer

Drive a Next.js route to instant navigation by setting up an agentic loop, under Cache Components / PPR, on initial load (hard navigation) and client-side navigation (soft navigation). Encode the goal as a failing @next/playwright instant() e2e and work it to green, one verified route at a time; the shipped test then…

vercel/next.js · 170 tokens