Harvest and Analyze

Collect neutral-atom QPU bitstrings by batch ID once a submission has run, correct them for detection error, compute the target observable, compare it to the emulated baseline from validate-emu, and return an accept/reject verdict. Runs after a submission, never instead of one. Triggered by phrases like "collect the QPU results", "harvest these batch IDs", "compare the QPU data to the emulation", "is the QPU data consistent with the noise model".

9 starsAI, LLMs & Agents Plugin marketplace

Install

Claude Code plugin
/plugin marketplace add pasqal-io/agent-marketplace

The repo is a Claude Code plugin marketplace. Run /plugin afterwards to pick the plugin that includes this skill.

skills CLI (Claude Code, Cursor, Codex and more)
npx skills add pasqal-io/agent-marketplace --skill harvest-and-analyze
Manual install
git clone --depth 1 https://github.com/pasqal-io/agent-marketplace.git /tmp/agent-marketplace
mkdir -p ~/.claude/skills
cp -r /tmp/agent-marketplace/skills/harvest-and-analyze ~/.claude/skills/harvest-and-analyze

Personal skills live in ~/.claude/skills. Use .claude/skills in a project folder to share a skill with your team.

From the SKILL.md

# harvest-and-analyze

Collect QPU results from the cloud, compute the observable, and compare to the
EMU baseline. The accept/reject decision is: does QPU agree with the noise model?

RUNS ON: this machine. It reads a finished submission — nothing is submitted and
no shots are bought here.

## Decisions that are not yours

- **the accept/reject criterion** — the σ tolerance, and what a disagreement is
  being read as (device drift, a wrong noise model, real physics)
- **whether to apply the detection correction** (Step 2b), which changes the
  number reported but must not change the verdict
- **what to conclude when only part of the scan completed**: report the partial
  curve, or wait for the queue

After about **three** attempts at the same failure — jobs that stay in `ERROR`, a
batch that will not read back — stop and report: which jobs, what status, what
is recoverable. Then offer waiting, re-submitting the missing points as their
own run, or analysing what did come back. Do not poll a dead batch in a loop.

---

## Support scripts

Paths like `support/…` below are relative to **this skill's directory** —
expand them to the skill's absolute location when running commands from your
project directory (keep outputs like `--out-dir` in your project, not the plugin).
Read the full skill on GitHub

Details

Skill name
harvest-and-analyze
Path
skills/harvest-and-analyze/SKILL.md
License
Apache-2.0
Repo updated
3 November 2026

FAQ

What does the Harvest and Analyze skill do?

Collect neutral-atom QPU bitstrings by batch ID once a submission has run, correct them for detection error, compute the target observable, compare it to the emulated baseline from validate-emu, and return an accept/reject verdict. Runs after a submission, never instead of one. Triggered by phrases like "collect the QPU results", "harvest these batch IDs", "compare the QPU data to the emulation", "is the QPU data consistent with the noise model".

How do I install the Harvest and Analyze skill?

Claude Code plugin: /plugin marketplace add pasqal-io/agent-marketplace. Skills go in ~/.claude/skills (personal) or .claude/skills (project), and Claude loads them automatically when a task matches the description.

Is Harvest and Analyze popular?

It lives in pasqal-io/agent-marketplace, which has 9 GitHub stars and 0 forks.

Related skills

Application to Idea

9

pasqal-io/agent-marketplace

Scope a rough problem or application into a first technical idea for a neutral-atom (Rydberg) QPU: which documented Pulser method could serve it — MIS, QUBO, an Ising or XY simulation, a graph kernel — what would be measured, at what size, or an honest no-fit. Writes a note for idea-to-spec; use idea-to-spec when a protocol or paper exists. Triggered by phrases like "can neutral atoms do anything with my problem", "is this a fit for a Rydberg QPU", "I don't know what to measure on the QPU".

Research & Science

Apply Genci Tgcc Cea

9

pasqal-io/agent-marketplace

Apply for QPU hours on Ruby (Pasqal Orion), the analogue neutral-atom machine at GENCI/CEA TGCC - via the French eDARI portal or the EuroHPC Quantum Access Pilot call - then get the TGCC computing account. For requests such as "apply for TGCC hours", "eDARI application", "apply for Ruby/Pasqal QPU allocation", "EuroHPC quantum hours", "Quantum Access Pilot", or "attach to a project / rattachement". Access only, not running experiments. Unsure which route fits? Use choose-access-method.

Software Development

Choose Access Method

9

pasqal-io/agent-marketplace

Decide which neutral-atom QPU access route to pursue - free CEA/GENCI/ EuroHPC access to Ruby (weeks of lead time, specific eligibility) versus Pasqal Cloud pay-as-you-go (today, any country, a card). For requests such as "how do I get QPU access", "which account should I use", "Pasqal Cloud vs CEA", "do I qualify for free QPU time", or "fastest way to real hardware". Hands off to create-pasqal-cloud-account or apply-genci-tgcc-cea: this skill is the decision, not the sign-up flow.

Software Development

Kamailio Config

8

UserAd/ClaudeSkills

Kamailio SIP server configuration and troubleshooting. Use when Claude needs to: (1) Edit or create Kamailio configuration files (.cfg) (2) Debug SIP routing logic (request_route, failure_route, reply_route) (3) Work with Kamailio pseudo-variables ($ru, $fU, $avp, $xavp, $dlg_var) (4) Configure modules (tm, dialog, acc, sqlops, jansson, http_client, rr, sl, mqueue, rtimer) (5) Understand SIP message flow and transaction handling (6) Validate configuration syntax Triggers: "kamailio", "SIP routing", "kamailio.cfg", SIP-related pseudo-variables

AI, LLMs & Agents

Prompt Forge

9

JordanGunn/nunchuck

Shape, refine, and stabilize human intent into a canonical prompt artifact. Iteratively clarifies ambiguity until user confirms readiness.

AI, LLMs & Agents

Tmux

8

UserAd/ClaudeSkills

Use tmux for orchestrating agents and running commands. Use when Claude needs to (1) spawn Claude subagents in background windows, (2) run long-running commands, (3) orchestrate parallel agents, (4) capture window output. Triggers on "tmux", "new tmux window", "run claude", "spawn agent", "parallel agents".

AI, LLMs & Agents

Browse the most popular Claude skills, all ai, llms & agents skills, or write your own with the skills generator.