A guided walk through the repo. About fifteen minutes if you run every step. About three minutes if you read along. You can do this yourself, or paste the prompt at the bottom into your agent and have it walk you through.
Start with three files at the repo root. Together they tell you the shape of everything else.
- README.md: elevator pitch, capability table, agent prompt, harness patterns.
- AGENTS.md: operating rules for any agent driving the repo. The skills, the schemas, the validators, the things to keep out of git.
- docs/glossary.md: cryo-EM vocabulary, tools, and CryoCore orchestration terms.
If you are new to cryo-EM, the glossary's "Cryo-EM Vocabulary" and "Cryo-EM Tools At A Glance" sections are the fastest domain bridge.
From the repo root:
python3 -m pip install -r requirements-dev.txt
make demo-localThat fetches public RCSB and mmCIF metadata for the T2R14 receptor complex
(PDB 9W0Q, EMDB EMD-65512), computes chain and ligand summaries, renders
SVG figures, and writes a review package under ignored .runtime/.
The demo runs in about a minute on a laptop CPU.
Open the headline artifacts:
ls .runtime/t2r14-open-dossier/artifacts/The three to look at first:
report.html: the human-readable review. Open it in a browser to see inputs, figures, and methods together.claim_ledger.md: what the evidence supports and where claims have to stop.dossier_manifest.json: the same content for an agent or downstream tool to read.
A static sample is committed at examples/t2r14-open-dossier-preview/ if you want to see the shape before running anything.
make release-checkThis runs the tests, the public-release report, the secret scan, and the contract validators. It is the same gate that CI runs. Green means the repo is publishable in its current state.
The individual checks are listed in docs/validation-command-matrix.md when you want a smaller, faster iteration loop.
The skill pack is what an agent actually loads when it works in this repo.
ls skills/
cat skills/cryocore/SKILL.mdEach skill has a SKILL.md (the human and agent-readable spec) and an agents/
directory with harness-specific configuration. skills/README.md
has a "Use When" table mapping each skill to the trigger that pulls it in.
The fastest way to put your agent to work is the paste-block in the Agent Prompt section of the README. After you paste it, hand the agent a goal in plain English. For example:
- "Build a public-accession map/model review for EMDB
EMD-43816and PDB9ASJ." - "Review the artifact evidence for a hypothetical Pol Theta run on RunPod."
- "Draft a Linear-style issue wave for a four-stage map-to-model campaign."
The agent reads AGENTS.md, picks the relevant skill, runs the validators, and
returns a review output or wave plan you can inspect.
| Track | Start here |
|---|---|
| Run the other two demos | Pol Theta, Structure Jury |
| Choose a workflow scale | Workflow Blueprints |
| Set up an agent harness | Agent Quickstart |
| Plan a Linear-style wave | Tracker Orchestration |
| Prepare a cloud or HPC lane | Compute Backends, Provider Readiness |
| Reuse the skill pack in another repo | Adoption Guide, Skill Installation |
| Browse copyable workflows | Recipes |
Paste this into your coding agent to have it walk you through the tour:
Use the CryoCore tour. Read docs/tour.md and follow each step in order. Run the
commands locally where the tour calls for them, summarize what you saw at each
step, and stop at any gate that asks for operator authorization. Report which
artifacts appeared, which validators passed, and which next track from the
table would be the smallest useful step from here.