mission-catalog.md

Mission Catalog

A menu of seed missions your coding agents can take on with CryoCore. Each entry names a starting point, the resource mode, the artifact that proves progress, and the rough scope. Pick one, hand it to your agent, and review what comes back.

The catalog is sorted from smallest to largest. Local-only missions need no network and no credentials. Public-metadata missions hit EMDB, PDB, and wwPDB. Provider missions stay in prep mode by default; a real paid run requires operator authorization.

Local-only missions

Mission	Resource mode	First artifact	Rough scope
Run make demo-local and write a one-paragraph summary of the review shape.	local	The summary itself, plus a pointer into .runtime/t2r14-open-dossier/.	15 minutes.
Compare two skills under skills/ and report when each one fires.	local	A short skill-comparison note in your agent's reply.	15 minutes.
Validate that every JSON file under modules/schemas/ and modules/campaigns/ parses cleanly.	local	make module-check output appended to a brief report.	15 minutes.
Audit references/software-registry.yaml for missing fields, stale versions, or broken posture entries.	local	A diff-or-note list of findings with suggested edits.	30 minutes.
Read docs/glossary.md and propose three new entries that would help a newcomer.	local	A short proposal block your agent posts back to you.	30 minutes.

Public-accession review missions

Mission	Resource mode	First artifact	Rough scope
Build a public-accession map/model review for an EMDB and PDB pair of your choice.	public-metadata-network	An input audit and figure-rendering plan that lists declared inputs and the claim ceiling.	1-2 hours.
Compare two public deposited structures of the same target and produce a state or structure-review plan.	public-metadata-network	A comparison plan citing both deposits and the difference axes the agent would compare.	1-2 hours.
Audit one PDB entry's wwPDB validation report and summarize what the deposit supports.	public-metadata-network	A bounded-claims note plus a short prose summary of the validation report.	1-2 hours.
Reproduce the Pol Theta map/model prep check and describe the contract-self-check output.	public-metadata-network	The prep-check log plus a contract-self-check pass record.	1 hour.

Advanced public-data missions

These are good second-wave tasks after the local demo works. They use public accessions and keep heavy files out of git.

Mission	Resource mode	First artifact	Rough scope
Plan an RNA map/model review for PDB 10ZT and EMDB EMD-75574. Focus on input inventory, figure plan, and which model-building tools would be tried first.	public-metadata-network	A public input ledger with EMDB, PDB, map, half-map, sequence, and wwPDB report links.	2-3 hours.
Plan a cryo-ET portal walkthrough for CryoET Data Portal dataset 10301. Focus on metadata intake, copick project shape, and what screenshots or annotation files an agent should return.	public-metadata-network	A dataset inventory and proposed copick config outline.	2-3 hours.

Tracker wave missions

Mission	Resource mode	First artifact	Rough scope
Draft a Linear-style wave for a four-stage raw-to-atomic campaign on EMPIAR-10204.	tracker-wave	Four issue bodies with dependencies, labels, and operator gates.	2 hours.
Plan a state-comparison wave that compares two structural interpretations of one public dataset.	tracker-wave	A wave plan with the comparison contract spelled out.	2 hours.
Re-shape the Pol Theta sample wave into one that runs the same checks against a different EMDB/PDB pair.	tracker-wave	Updated issue bodies pointing at the new accessions.	1 hour.

Provider missions (prep mode)

Mission	Resource mode	First artifact	Rough scope
Prepare a RunPod bridge manifest for a public-accession map/model review and confirm make runpod-scope-check and make runpod-reference-check pass.	operator-gated-provider (prep only)	Validated bridge manifest, scope report.	1-2 hours.
Audit one provider profile under modules/provider-profiles/ and propose updates to its launch-request template.	operator-gated-provider (prep only)	A diff or note listing proposed updates with justifications.	1-2 hours.
Draft a stage contract for a map-to-model lane on a generic cloud VM.	operator-gated-provider (prep only)	New stage-contract JSON under runpod/stage-contracts/ with passing validators.	2 hours.

Provider run review missions

Mission	Resource mode	First artifact	Rough scope
Review a hypothetical RunPod artifact package and decide whether it should be treated as complete.	provider-closeout	A pass/blocker report referencing artifacts, hashes, cost records, and cleanup proof.	1 hour.
Score the contract-self-check output of one fixture run and explain any blockers in plain English.	local	A short report mapping each blocker to the contract field it failed.	30 minutes.

Skill-pack and adoption missions

Mission	Resource mode	First artifact	Rough scope
Adapt CryoCore patterns into a new structural-biology repo (X-ray, NMR, or hybrid).	local	A starter SKILL.md and adoption notes pointing at the parts that need domain re-wiring.	1 day.
Add a new skill under skills/<your-skill>/SKILL.md for a workflow that is currently missing from the skill pack. Run make skill-check.	local	New skill directory with passing validators and an entry in skills/index.yaml.	1 day.
Copy the skill pack into another scientific repo and report which validators carry over without changes.	local	A short adoption report with the carry-over list and the gaps.	1 day.

Documentation and polish missions

Mission	Resource mode	First artifact	Rough scope
Add a glossary entry for a cryo-EM term that is currently missing.	local	Updated docs/glossary.md with the new entry and links from related docs.	30 minutes.
Expand a recipe in docs/recipes/ with concrete copy-paste flows and expected outputs.	local	Updated recipe doc and a passing make docs-link-check.	1 hour.
Add a new agent-task prompt fixture under examples/agent-tasks/ for a use case not yet covered.	local	New .prompt.md and an entry in examples/agent-tasks/README.md.	1 hour.

How to use the catalog

Pick the smallest mission that captures the work you actually want done. Paste the row's mission text into your agent along with the Agent Prompt from the README. The agent will read the relevant skill, run the validators named in the linked docs, and report what it found.

If a mission ends up larger than expected, your agent should split it into a tracker wave using the Linear Wave Planning prompt and stop at the first operator gate.