sm-analyses - SKILL.md Agent Skill

name: sm-analyses description: Use when the user names a Standard Model process and wants to "rediscover" it with the public ATLAS Open Data — Z→ll, H→ZZ→4l, H→γγ, H→μμ, H→bb, t-tbar, or WZ→3l+ν. Pairs the process to its docs section in `13TeV25Doc/StandardModel` and its notebook in `notebooks-collection-opendata/13-TeV-examples/uproot_python/`. Does NOT cover BSM searches, generic notebook indexing by filename (use `atlas-notebooks`), or analysis-code authoring (use `physlite-basics` or the `analysis` agent). Disambiguator phrase: SM walkthrough by process. data_scope: open experiment: atlas

Scope

Use this skill for the SM walkthrough chapters at https://opendata.atlas.cern/docs/13TeV25Doc/StandardModel and their corresponding notebooks in atlas-outreach-data-tools/notebooks-collection-opendata/13-TeV-examples/uproot_python/.

If the user has a different process (e.g. a BSM search, a ML example, or a technical tutorial), use atlas-notebooks instead.

Walkthrough table

Process	Docs section	Notebook
Z → μμ (and Z → ee / other SM extensions)	Find the Z boson	`Find_the_Z.ipynb`
H → ZZ* → 4ℓ	Higgs to ZZ	`HZZAnalysis.ipynb`
H → γγ	Higgs to γγ	`HyyAnalysis.ipynb`
H → μμ	Higgs to Muon–Antimuon	`Hmumu.ipynb`
H → bb	Higgs to bb	`HbbAnalysis.ipynb`
tt̄	Top–Antitop	`ttbar_analysis.ipynb`
WZ → 3ℓ + ν	WZ to three leptons	`WZ3l_Analysis.ipynb`

All seven notebooks live in 13-TeV-examples/uproot_python/ of the outreach repo.

Common procedure

For any of the seven walkthroughs:

Read the docs section (quote the URL). It gives the physics motivation, the event selection, and the expected plot.
Open the matching notebook (13-TeV-examples/uproot_python/<file>.ipynb) in the user's runtime of choice (via the atlas-notebooks skill).
Identify the MC samples the notebook needs. Use atlas_match_metadata with keywords / processes from the docs (e.g. Zee, Zmumu, ttbar, Higgs, ZZ, WZ).
Confirm the release via atlas_get_current_release — the 2025 SM docs correspond to the 2024r-pp or 2025r-evgen-13tev releases; use the one the notebook header specifies.
Reproduce the expected plot. Compare the position of the peak with the known mass (91.2 GeV Z, 125 GeV H, 172.5 GeV t).

Per-walkthrough notes

Find_the_Z (Z → μμ)

Selection: 2 opposite-sign muons, dimuon invariant mass. Expected peak at ~91.2 GeV. Extensions suggested in the docs: Z→ee, J/ψ, Υ by changing the mass window.

HZZAnalysis (H → ZZ* → 4ℓ)

Selection: 4 leptons, 2 same-flavour opposite-sign pairs, m_Z1 near 91 GeV, m_Z2 in [12,60] GeV. Expected narrow peak at m_4l ≈ 125 GeV. Companion coffea-based version: CoffeaHZZAnalysis.ipynb. The ML demos (HZZ_BDT_demo.ipynb, HZZ_NeuralNet_demo.ipynb) build on this output.

HyyAnalysis (H → γγ)

Uses 36.1 fb⁻¹ (per the docs). Selection: 2 isolated photons with tight ID, m_γγ in [105,160] GeV. Fit the m_γγ spectrum with a continuum background + Gaussian signal at 125 GeV. RDataFrame variants live in 13-TeV-examples/rdataframe/RDataFrame_Hyy.ipynb and RDataFrame_Hyy_SimpleExample.ipynb.

Hmumu (H → μμ)

Very suppressed branching ratio; the notebook teaches signal extraction in a large background. Peak at 125 GeV on a steeply falling DY+jets continuum.

HbbAnalysis (H → bb)

Advanced — the docs flag this as attempting to find H → bb. Uses b-tagged jets and requires MVA techniques. Not a good first walkthrough.

ttbar_analysis (tt̄)

Lepton + jets channel. Selection: 1 lepton, ≥4 jets with ≥2 b-tags, MET. Expected shape in jet multiplicity and b-tag multiplicity.

WZ3l_Analysis (WZ → 3ℓν)

Selection: 3 leptons, one opposite-sign same-flavour pair near 91 GeV (the Z), remaining lepton + MET forms the W transverse mass.

Pitfalls

Notebook headers sometimes reference older releases (e.g. 2020e-13tev). Check with atlas_get_current_release and atlas_set_release before pulling sample metadata — DSIDs change across releases.
The ML demos (HZZ_BDT_demo, HZZ_NeuralNet_demo) are not teaching ML from scratch; they assume the user already has the HZZAnalysis.ipynb ntuples.
Hbb is not a clean "rediscovery" in the outreach release; don't over-promise a peak.
Luminosity numbers: 36.1 fb⁻¹ for older Hyy; ~140 fb⁻¹ for Run 2 full. Check the docs section, not the notebook, to get the authoritative number.

Verification

A successful use of this skill ends with the user:

Open on the correct docs page and matching notebook.
Aware of the MC samples to query (via atlas_match_metadata).
Reproducing the expected peak / distribution within a reasonable tolerance of the docs figure.