re-com: Skill for Claude Code

.claude/skills/recom-quality-check/SKILL.md

recom-quality-check is a skill for Claude Code from umsachde/re-com. It costs 56 tokens per session (888 once invoked), scanned A, original, MIT.

A procedure for measuring the quality of re-com’s recommendations and ranking changes. It uses metrics such as mood fit, overlap, similarity agreement, and run-to-run variation.

In plain words
What is it for?
It is for evaluating recommendation results, comparing ranking versions, and checking whether changes improve or harm scoring quality.
Why use it?
It replaces judging ranking changes by impression alone and helps distinguish a real improvement from normal measurement noise.

Skill for Claude Code

Written for Claude Code: installed under .claude/.

This is umsachde/re-com's own configuration. It tells Claude Code how to work on re-com itself, so it is not a mod to install elsewhere. Copy it as a starting point and replace the rules that are about this project. Everything re-com configures →

Needs its repository: it runs a file that does not travel with it, so clone the repository first. The line is python scripts/quality_check.py --titles # mood/arc cases, human-readable output.

Reuse

Borrowing it

Nothing to install: this file belongs to umsachde/re-com. Take a copy, put it at the same path in your own repository, and replace the rules that are about this project with yours.

Copy the file
curl -O https://raw.githubusercontent.com/umsachde/re-com/main/.claude/skills/recom-quality-check/SKILL.md
Clone the repo
git clone --depth 1 https://github.com/umsachde/re-com

Made for: Claude Code.

Wrote 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.

agentmods badge for recom-quality-check

README.md
[![agentmods](https://agentmods.dev/badge/skills/umsachde/re-com/recom-quality-check/github.svg)](https://agentmods.dev/skills/umsachde/re-com/recom-quality-check)
Your own site
<a href="https://agentmods.dev/skills/umsachde/re-com/recom-quality-check"><img src="https://agentmods.dev/badge/skills/umsachde/re-com/recom-quality-check/github.svg" alt="Measured on agentmods" height="20"></a>

Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.

agentmods 80×15 button for recom-quality-check

Your own site · 80×15
<a href="https://agentmods.dev/skills/umsachde/re-com/recom-quality-check"><img src="https://agentmods.dev/badge/skills/umsachde/re-com/recom-quality-check.svg" alt="Reviewed on agentmods" width="80" height="20"></a>
Per session 56 Skills are progressive disclosure: only the name and description are preloaded; the body loads when the skill is used.
When invoked 888 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. A grade says what 26 rules found in the file — not that it is safe.
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 5.1 $0.00056 $0.00888
Opus 5.5 $0.00022 $0.00355
Sonnet 5.5 $0.00011 $0.00178
Haiku 4.5 $0.00006 $0.00089

Measured 2mo ago against content hash 459ef489fa33, method: parsed. Prices are Anthropic first-party input rates as of 2026-10-07, from the pricing page.

Security

Grade A, and why

recom-quality-check 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 2mo 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.

.claude/skills/recom-quality-check/SKILL.md · 69 lines

How it starts

The opening of the file, as written. The whole thing — 69 lines — stays where its author put it; the contents beside it link to each section on GitHub.

scripts/quality_check.py is the third verification layer (PLAN.md §5) — it turns ranking-quality questions into numbers instead of impressions. Run it by hand whenever a change touches ranking, signal weighting, mood scoring, or the graph.

Commands

python scripts/quality_check.py --titles                    # mood/arc cases, human-readable output
python scripts/quality_check.py --distinctiveness 0          # A/B the seed-scoring logic
python scripts/quality_check.py --similarity --repeat        # scores recommend_from_song / recommend_from_playlist, plus a noise floor in the same run

Always prefer --repeat when comparing two versions of the ranking logic — it measures the noise floor (run-to-run variance) in the same invocation, so a delta can be judged against it instead of against a stale number from PLAN.md.

The metrics, and what they actually mean

  • Mean mood fit — average fit score. Do not stop here. PLAN.md §3 records a build that scored a healthy 0.775 mean fit while returning 70% the same songs for "heartbroken" and "angry" — fit alone cannot see that kind of collapse.
  • Cross-mood overlap (lower is better) — the metric that does catch that failure mode. Always check this alongside mean fit, never instead of it.
  • Cross-seed overlap (lower is better, --similarity only) — the same-shape check for the similarity path: whether many different seeds funnel into one popular attractor.
  • Signal agreement / corroborated (--similarity) — reported against a per-backend ceiling, not as a raw count. YouTube's ceiling is higher than Spotify's because Spotify has fewer native signals (capabilities() is emptier there) — a bare cross-backend mean would misread that difference as a regression.
  • Native-vs-graph A/B: churn vs. corroboration delta — churn alone cannot answer "did this help or dilute", because both arms truncate to the same limit and can show identical churn/additions by pure arithmetic (PLAN.md §6.5 — this happened, 37 == 37). The corroboration delta is the number that actually answers helping-vs-diluting.
  • Artist concentration (HHI) — measure it before any max_per_artist cap is applied; measuring after only confirms the cap works, it doesn't tell you anything about the underlying ranking.

Read the full file on GitHub · 69 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. 2mo ago First seen · 69 lines · 56 tokens per session scan A 459ef489fa33

Subscribe to this mod's changes

recom-quality-check is a skill published in the GitHub repository umsachde/re-com (0 stars, last pushed 1mo ago), licensed MIT. It adds 56 tokens to every session and 888 once invoked, about $0.0002 per session on Opus 5.5. 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 2026-09-14.

Related

recom-quality-check compared with skills from other repositories

recom-quality-check next to 6 skills from other repositories
ModTypeStarsLast pushLicenceScanTokensWorks with
recom-quality-checkumsachde/re-comSkill01mo agoMITA56Claude Code
imagegenopenai/codexSkill136ktodayApache-2.0A113Codex
local-ai-agentsmicrosoft/ai-agents-for-beginnersSkill80kyesterdayMITA200Claude Code, Codex
playwright-climicrosoft/playwrightSkill100ktodayApache-2.0A19Claude Code
browser-automationopenclaw/openclawSkill395ktodayMITA36Claude Code
free-llm-apismnfst/awesome-free-llm-apisSkill14ktodayCC0-1.0A113Claude Code, Codex
next-cache-components-adoptionvercel/next.jsSkill144ktodayMITA95Claude Code