New: 98,000+ skills indexed from GitHub repos with 100+ stars · every skill safety-checkedBrowse top skillsTop skills
Esc to closeBrowse all skills
Submit a skillBrowse skills
Submit a skillBrowse skills

Agent skill

Paro Optimizer

by zunor in zunor/paro

Design, refactor and diagnose Paro's staged optimizer, using EXPLAIN COMPILE for planning and EXPLAIN ANALYZE for execution.

Apache-2.0Auto-check passedDatabases

Install Paro Optimizer

skills CLI
$ npx skills add zunor/paro --skill paro-optimizer -a claude-code

Project install by default; add -g for ~/.claude/skills/.

GitHub CLI
$ gh skill install zunor/paro paro-optimizer --agent claude-code

Project scope by default; add --scope user for a personal install. Needs GitHub CLI 2.90.0 or later (public preview).

Manual copy
$ git clone --depth 1 https://github.com/zunor/paro.git skills-src && mkdir -p .claude/skills && cp -r skills-src/.agents/skills/paro-optimizer .claude/skills/paro-optimizer && rm -rf skills-src

Use ~/.claude/skills/ instead of .claude/skills for a personal install. The folder must contain SKILL.md.

Claude Code skills documentation · loads skills from .claude/skills/

Facts

Skill name
paro-optimizer
GitHub stars
105
Token cost
~979 tokens
SKILL.md length
437 words
Files
2
Skills in repo
3
Repo updated
First seen
Licence
Apache-2.0

At a glance

Design, refactor and diagnose Paro's staged optimizer, using EXPLAIN COMPILE for planning and EXPLAIN ANALYZE for execution.

  • Works in 5 steps: Check estimated versus actual rows at… → Use EXPLAIN (COMPILE, DETAIL, FORMAT… → For a costly choice, inspect both… → …
  • Optimizer changes and plan-quality investigations
  • SKILL.md covers Place the change with its owner, Diagnose before changing a… and Validate proportionally
  • Instructions only: no scripts, shell commands, URLs or credentials in SKILL.md

What it does

Paro Optimizer is an agent skill from zunor/paro. Design, refactor and diagnose Paro's staged optimizer, using EXPLAIN COMPILE for planning and EXPLAIN ANALYZE for execution. Use for optimizer changes and plan-quality investigations.

Its SKILL.md is about 980 tokens, which your agent loads only when the skill is triggered. The skill folder holds 2 other files (for example `agents/openai.yaml`).

It sits in Databases, covering Query optimization and Vector databases. It works with SQL, PostgreSQL and Rust. The repository describes itself as: An AI-native multi-model database unifying SQL, vector, full-text, graph, and sandboxed Python — for transactional, analytical, and agent workloads. The licence is Apache-2.0.

When your agent uses it

  • Optimizer changes and plan-quality investigations
  • Tasks that involve Query optimization
  • Tasks that involve Vector databases

Example prompts

  • “/paro-optimizer”

Workflow steps

5 steps, taken from the first numbered list in SKILL.md.

  1. Check estimated versus actual rows at the first divergence, including CTE
  2. Use EXPLAIN (COMPILE, DETAIL, FORMAT JSON) for stage work and bounded
  3. For a costly choice, inspect both alternatives' rows, widths, work and
  4. Prioritize corpus excess-time/outlier rankings, not only Q04/Q11/Q74.
  5. Measure first-execution phases directly. Differences of cohort medians

What it can do on your machine

Read from SKILL.md and the folder at commit 4afe117. It shows what the files ask for, not the result of running them.

  • Tool permissions

    Pre-approves nothing: there is no allowed-tools line, so your agent's usual permission prompts apply.

    From allowed-tools in the SKILL.md frontmatter.

  • Runs code

    No scripts in the folder and no shell commands in SKILL.md.

    From the folder's file list and the shell code blocks in SKILL.md.

  • Network

    No URLs in SKILL.md.

    From URLs in SKILL.md, links to its own repository left out.

  • Credentials

    Names no API keys, tokens, secrets or passwords.

    From names ending in _API_KEY, _TOKEN, _SECRET, _KEY or _PASSWORD in SKILL.md.

Context cost

Paro Optimizer loads about 979 tokens when it runs. Until then it costs about 50 tokens; SKILL.md has 437 words of instructions outside code blocks.

Always · name and description, kept in context so the agent knows when to use it
~50
When it runs · the whole SKILL.md, loaded when a task matches
~979

Estimates: characters ÷ 4, the usual rule of thumb; real counts depend on the model's tokenizer. Scripts and assets cost tokens only if the agent reads them.

Safety

Auto-check passed

The automated check found no risky patterns in SKILL.md.

Automated static check — not a guarantee. Review scripts before installing. It scans the text of SKILL.md for risky patterns (piping downloads into a shell, reading credential files, hidden Unicode, destructive commands); files beside SKILL.md are not scanned.

SKILL.md

The full file from zunor/paro at commit 4afe117, republished under its Apache-2.0 licence (© zunor). 437 words, ~979 tokens.

Download SKILL.mdSave it as .claude/skills/paro-optimizer/SKILL.md (or your agent's skills folder). This skill also uses 1 other file; get the full folder from GitHub.
name
paro-optimizer
description
Design, refactor and diagnose Paro's staged optimizer, using EXPLAIN COMPILE for planning and EXPLAIN ANALYZE for execution. Use for optimizer changes and plan-quality investigations.

Paro optimizer

Start with the selected checkout's HEAD/status and architecture. Preserve unrelated work. Keep one production planner, not alternative policies.

Place the change with its owner

  • rewrite::program: ordered semantic replacements, without cost alternatives.
  • estimate::annotate: shared column/relation/selectivity kernels; distinguish expected estimates, proven bounds and unknowns.
  • region::plan: interacting join/aggregate decisions, bounded enumeration. Transitions consume compact summaries, not cloned plans or full physical selection.
  • physical::choose / lower: local algorithms, access paths, RF and committed physical construction. Cost assumptions must match emitted keys/residuals.
  • paro-planner: binding and shared logical/physical contracts. Execution consumes plans, not optimizer internals. Runtime adaptation belongs in execution and must respect memory, spill, cancellation and output contracts.

One concept has one owner. Keep production kernels and their tests; remove uncalled alternative algorithms instead of retaining them as test-only “oracles”.

Diagnose before changing a decision

  1. Check estimated versus actual rows at the first divergence, including CTE boundaries, complete keys, NULL semantics and predicate activation. More enumeration does not repair wrong estimates.
  2. Use EXPLAIN (COMPILE, DETAIL, FORMAT JSON) for stage work and bounded regional decisions; use EXPLAIN ANALYZE for actual operator behavior. Read compile diagnostics and the actual typed schema, not old report names.
  3. For a costly choice, inspect both alternatives' rows, widths, work and feasibility. A scoped forced-choice experiment can isolate its effect; don't ship query-specific join/build hints or fit constants to one query.
  4. Prioritize corpus excess-time/outlier rankings, not only Q04/Q11/Q74. A measured local decision can justify runtime adaptation; “unknown” alone does not make an unimplemented adaptive path safe.
  5. Measure first-execution phases directly. Differences of cohort medians give scale, not a causal breakdown.

Planned / PlannedWithFallback describe legal artifact production, not global optimality. Fingerprints associate plans; they do not prove SQL equivalence. Never pair normal timings with an unrelated diagnostic plan.

Show full SKILL.md (146 more words)Show less

Validate proportionally

  • Mechanical refactors: preserve public contracts, typed identity and EXPLAIN body; for broad planner changes compare TPC-H 22 / TPC-DS 99 before/after, full typed results, and SQL regress. Explain expected differences explicitly.
  • Plan-changing work: counterexamples first (NULLs, duplicates, evaluation errors, outer/recursive boundaries), then real-entry tests and corpus results; inspect excess-time rankings for new outliers.
  • Use paro-benchmark for ordinary performance exploration. Formal non-inferiority/parity uses paro-evidence, not every edit. No failed semantic comparison may be hidden by snapshot regeneration.

For source comparisons, reuse the agreed small worktree and one shared Cargo target sequentially; save binary/source identities before rebuilding. Do not create large targets/data copies automatically. Reuse immutable, relocatable seeds. Normal compile timing needs the checkout's bounded compile-work observer (PARO_COMPILE_WORK_EVIDENCE=1 in the collector's environment); verify an actual receipt's compiler_elapsed_us before collecting more samples. Missing compile data is uncovered, not zero, and Detail time is not a normal compile sample.

© zunor, Apache-2.0. Rendered from Markdown: HTML in the file is shown as text, images as links, and headings moved down two levels. Raw file

Files

SKILL.md and 1 other file in .agents/skills/paro-optimizer of zunor/paro.

  • SKILL.md
  • agents/openai.yaml

Open the folder on GitHubat commit 4afe117

Compare with similar skills

Paro Optimizer next to the 5 skills that share the most tags, products or categories with it. Stars are the repository's; “used in” counts other GitHub owners with a copy.

Paro Optimizer compared with similar skills
SkillStarsUsed inTokensAuto-checkLicenceRepo updated
Paro Optimizer this skillzunor/paro105—~979Automated safety check: PassApache-2.0
PostgreSQL Documentation Reference2025Emma/vibe-coding-cn23k1 repos~19kAutomated safety check: PassMIT
Diesel Guardayarotsky/diesel-guard121—~3.1kAutomated safety check: PassMIT
Veloxdb Scalable Performanceveloxbase/veloxdb646—~1.7kAutomated safety check: PassMIT
Querying Tempotempoxyz/tidx107—~3.1kAutomated safety check: PassMIT
Neon Postgresneondatabase/agent-skills100—~4.1kAutomated safety check: NotesApache-2.0

Similar skills

  • PostgreSQL Documentation Reference

    2025Emma/vibe-coding-cn

    PostgreSQL database documentation - SQL queries, database design, administration, performance tuning, and advanced features. Use when working with PostgreSQL…

    23k GitHub starsUsed in 1 repo~19k tokens
    DatabasesAuto-check passed

  • Diesel Guard

    ayarotsky/diesel-guard

    Lints Diesel and SQLx Postgres migrations for unsafe schema changes that lock tables or cause downtime, and authors custom Rhai checks.

    121 GitHub stars~3.1k tokensUpdated 10 days ago
    DatabasesAuto-check passed

  • Veloxdb Scalable Performance

    veloxbase/veloxdb

    Guides scalability and performance work for VeloxDB's Tauri + Rust PostgreSQL backend and React + TanStack frontend.

    646 GitHub stars~1.7k tokensUpdated 8 days ago
    DatabasesAuto-check passed

  • Querying Tempo

    tempoxyz/tidx

    Query indexed Tempo chain data via tidx HTTP API and CLI. An agent skill from tempoxyz/tidx.

    107 GitHub stars~3.1k tokensUpdated yesterday
    DatabasesAuto-check passed

  • Neon Postgres

    neondatabase/agent-skills

    Official

    Guides and best practices for working with Lakebase Postgres on Neon: connections, pooled vs direct, schema migrations, branching, autoscaling, scale-to-zero, instant restore, read replicas, IP…

    100 GitHub stars~4.1k tokensUpdated 2 days ago
    DatabasesAuto-check: notes

  • DB Sculptor

    EliasOulkadi/shokunin

    Design database schemas with Prisma/Drizzle, PostgreSQL index strategy (B-tree, GIN, GiST, BRIN, Hash), query optimization (EXPLAIN ANALYZE), migration safety (expand/contract, zero-downtime), and…

    114 GitHub stars~3.1k tokensUpdated 4 days ago
    DatabasesAuto-check: notes

More from zunor/paro

  • Paro Benchmark

    zunor/paro

    Run Paro engineering performance gates and exploratory cold/warm, cross-engine or operator comparisons.

    105 GitHub stars~1.8k tokensUpdated yesterday
    Auto-check passed

  • Paro Evidence

    zunor/paro

    Design or audit formal Paro performance claims: release acceptance, cross-engine parity, model calibration certification or non-inferiority.

    105 GitHub stars~1k tokensUpdated yesterday
    Auto-check passed

Works with

Categories

Questions about Paro Optimizer

What does Paro Optimizer do?

Design, refactor and diagnose Paro's staged optimizer, using EXPLAIN COMPILE for planning and EXPLAIN ANALYZE for execution. Paro Optimizer is an agent skill from zunor/paro. Design, refactor and diagnose Paro's staged optimizer, using EXPLAIN COMPILE for planning and EXPLAIN ANALYZE for execution.

When should I use Paro Optimizer?

Paro Optimizer fits situations like: optimizer changes and plan-quality investigations; tasks that involve Query optimization; tasks that involve Vector databases.

How do I install Paro Optimizer in Claude Code?

Run `npx skills add zunor/paro --skill paro-optimizer -a claude-code`. Or copy the skill folder (.agents/skills/paro-optimizer in zunor/paro) into .claude/skills/paro-optimizer in your project. Claude Code loads it when a task matches its description.

How do I install Paro Optimizer in Codex?

Run `npx skills add zunor/paro --skill paro-optimizer -a codex`. Or copy the skill folder (.agents/skills/paro-optimizer in zunor/paro) into .agents/skills/paro-optimizer in your project. Codex loads it when a task matches its description.

Can I use Paro Optimizer in Cursor, Gemini CLI or GitHub Copilot?

Cursor, Gemini CLI, GitHub Copilot and OpenCode also load SKILL.md folders. With the skills CLI, run `npx skills add zunor/paro --skill paro-optimizer -a cursor` (or -a gemini-cli, github-copilot or opencode for the others). To copy it by hand, put the folder in .cursor/skills/paro-optimizer, .gemini/skills/paro-optimizer, .github/skills/paro-optimizer and .opencode/skills/paro-optimizer in your project.

What does Paro Optimizer need to run?

SKILL.md names no scripts, command-line tools or credentials: Paro Optimizer is instructions for the agent only.

Does Paro Optimizer access the network?

SKILL.md contains no URLs. Any network use would come from the scripts or tools the agent runs. This is read from the text; nothing was executed.

Is Paro Optimizer safe to install?

Our automated static check of SKILL.md found no risky patterns, such as piping downloads into a shell, reading credential files or hidden Unicode. It is not a guarantee. Review the folder before installing.

What licence does Paro Optimizer use?

Paro Optimizer is published under the Apache-2.0 licence (the repository's licence). It allows redistribution, so the full SKILL.md is shown on this page.

How many tokens does Paro Optimizer use?

About 979 tokens (SKILL.md is roughly 3.9k characters). Agents keep only the skill's name and description in context until a task matches; then they load SKILL.md in full.

What are the alternatives to Paro Optimizer?

Skills that share tags, products or a category with Paro Optimizer: PostgreSQL Documentation Reference (2025Emma/vibe-coding-cn, 23k stars), Diesel Guard (ayarotsky/diesel-guard, 121 stars), Veloxdb Scalable Performance (veloxbase/veloxdb, 646 stars) and Querying Tempo (tempoxyz/tidx, 107 stars). The comparison table on this page puts their stars, adoption, token cost, safety result and licence side by side.

Who maintains Paro Optimizer?

zunor (a GitHub user) maintains it in zunor/paro, which has 105 GitHub stars. The repository holds 3 skills in this directory. The repository was last updated on October 8, 2026.

Source: zunor/paro on GitHub. Facts on this page come from the repository at the commit we read; the author's words are quoted as theirs.