Use when the user needs authenticated Rucio data-management queries — DIDs, RSEs, replicas, metadata, replication rules — on collaboration-internal data, via EITHER the remote `rucio-*` MCP tools (rucio-atlas, rucio-cms, rucio-escape, rucio-fcc; per-user SSO, no local client needed) OR the Rucio v38+ CLI on lxplus, SWAN, or a CVMFS-staged client (`/cvmfs/sw.escape.eu/rucio/<version>/`). Read queries default to the MCP tools; `rucio download`/`upload` and rule writes require the CLI. CLI is ALWAYS noun-verb (`rucio rse list`, `rucio did show`, `rucio rule add`); NEVER the deprecated flat verbs (`list-rses`, `list-dids`, `add-rule`, `get-metadata`, `rule-info`) even though MCP tool names (`rucio_list_dids`) resemble them. Targets non-public datasets — for ATLAS Open Data DSIDs use `atlas-opendata` instead; for grid submission use PanDA, not Rucio directly. Disambiguator phrase: rucio v38 noun-verb.
Use when the user has an ATLAS DSID, `physics_short` name, or ATLAS Open Data release tag (`2024r-pp`, `2025r-evgen-13tev`) and needs file URLs, cross-sections, k-factors, filter efficiencies, sumOfWeights, or MC weight metadata. Backed by the `atlasopenmagic` MCP. Operates on ATLAS Open Data only — for collaboration-internal datasets use `rucio`; for non-ATLAS portal records identified by `recid` or DOI use `cern-opendata`; for configuring or running an event generator itself — Sherpa steering files → `sherpa-manual`, MadGraph `.mg5` scripts / cards → `madgraph` — use that skill; `atlas-opendata` only resolves metadata for already-produced samples, including Sherpa- (`Sh_…`) and MadGraph- (`MG_…` / `aMC…`) produced ones. Disambiguator phrase: atlasopenmagic DSID resolver.
Use when the user describes a goal at the CERN/HEP infrastructure level and wants to know WHICH services to stitch (e.g. "I want distributed analysis on Open Data and to train an ML model"). Returns a 2–4-service stack with steps and pointers to the matching execution skill (`rucio`, `reana-workflows`, `physlite-basics`, …). Bundles a service catalogue (`reference/catalog.yaml`), pre-cooked recipes (`reference/recipes.md`), and digests on GPU access, SWAN+HTCondor scale-out, and columnar frameworks. Does NOT cover physics methodology, single-tool command-level help, or running specific code (use the dedicated skill named by the recommendation). Disambiguator phrase: CERN service stack composer.
Use when the user wants a multi-reviewer critique of an analysis ARTIFACT — a notebook, an analysis-note PDF or text, or an analysis repo — judged for physics correctness, plot validity, and unmet commitments, ending in a single PASS / ITERATE / ESCALATE verdict. Defines the panel protocol: who reviews (the `reviewer-physics`, `reviewer-critical`, `reviewer-constructive` agents and the `plot-validator` skill), the A/B/C finding classes, panel composition per artifact type, the 3/5/10 iteration caps, the fresh-reviewer rule, and how the `arbiter` adjudicates. Does NOT gate a completion claim for a long-running compute step (that is `verification-before-completion`, the completion-evidence gate); does NOT extract numbers from one paper (that is `read-publication`); does NOT itself run the figure checks (that is `plot-validator`). Disambiguator phrase: A/B/C review-panel protocol.
Use to run the deterministic red-flag checks on an analysis FIGURE — a histogram PNG/PDF or its histogram/fit JSON — before a human or a reviewer agent opines on it. The shipped `scripts/plot_checks.py` checks the file exists and is non-empty, data/MC ratios sit inside a profile's band, no empty/negative bins, the fit has positive degrees of freedom (catches `chi2/ndf = 0/0`), goodness-of-fit is acceptable, and the MeV/GeV axis-unit trap; thresholds come from `reference/thresholds.md`, not the code. Then VIEW the rendered image (PDF → text via the AGENTS.md pdftotext/pypdf recipe). NOT the multi-reviewer protocol that decides PASS/ITERATE/ESCALATE (that is `analysis-review`); NOT the compute-step completion gate (that is `verification-before-completion`). Disambiguator phrase: programmatic plot red-flag checks.
Use when the user asks for a canonical Particle Data Group value — particle mass, lifetime, branching ratio, decay width, magnetic moment, mixing parameter, charge, spin, or any quoted "PDG average" constant. Backed by https://pdg.lbl.gov/ via WebFetch (HTML pdgLive pages and the PDG REST API where available). Always cites the PDG record URL and edition year. Does NOT cover ATLAS Monte Carlo metadata (use `atlas-opendata`), measured values from a specific paper (use `read-publication`), HEPData tabulated measurements (use `hepdata`), or conceptual physics explanations of why a particle has the value it does (use the `tutor` agent). Disambiguator phrase: PDG particle data group lookup.
Use when the user wants to submit, monitor, inspect, or kill HTCondor batch jobs at CERN — `condor_submit` against a `.sub` file, `condor_q` for status, `condor_history` for finished jobs, `condor_rm` to kill, `condor_status` for pool health — plus the `htcondor` / `htcondor2` Python bindings. Covers `+JobFlavour` wall-time picks, multi-CPU / GPU requests, automatic Kerberos / AFS / EOS credential handling on lxplus / aiadm submit hosts. Detect availability via `command -v condor_submit`. Does NOT cover REANA workflow runs (use `reana` for status / logs, `reana-workflows` for authoring), CERN batch documentation lookup (use `cern-docs` with `source=batch`), proof that a finished job actually produced the expected outputs (use `verification-before-completion`), or multi-service infra recommendations (use `infra-advisor`). Disambiguator phrase: HTCondor lxbatch CLI & Python submission.
Use when the user needs to read or write HepMC3 event records in Python — generator-level event graphs, vertices, particle production chains — with the `pyhepmc` package. Does NOT cover Les Houches Event files (use `pylhe`), detector-level ROOT ntuples (use `physlite-basics`), or 4-vector math on the particles (use `vector`). Disambiguator phrase: pyhepmc HepMC3 event records.