Bowen lab in loop planner
Skill allenlee0430/bowen-ai-skills/skills/bowen-lab-in-loop-planner
Reusable Codex and Claude Code skills for research operations, grants, evidence synthesis, lab-in-the-loop planning, file curation, and faculty workflows.
npx -y skills add allenlee0430/bowen-ai-skills --skill bowen-lab-in-loop-plannerAssembled from the repository path, not quoted from the project. Check it against their README if it does not work.
One thing to look at
- 0 stars0 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.
What its author says it does
Copied from the file, not written here
Design lab-in-the-loop and self-driving research programs for LNPs, biomaterials, RNA therapeutics, vaccines, and gene editing. Use when converting a therapeutic objective into a closed-loop design-build-test-learn architecture with explicit variables, assays, data standards, uncertainty-aware model decisions, stage gates, validation, and stop criteria.
SKILL.md
1.7 KB, 247 tokens by cl100k_base, as published. Nobody here has run it
Bowen Lab-in-the-Loop Planner
Workflow
- Read
references/program-design.md. - Define the target product profile and the biological decision the program must improve.
- Separate design variables, nuisance variables, objectives, hard constraints, and validation endpoints.
- Create a staged assay funnel. Assign each model and assay one explicit role; do not ask a high-throughput proxy to prove clinical relevance.
- Specify the data contract: entity identifiers, formulation, process parameters, assay context, controls, replicates, QC, missingness, provenance, and negative results.
- Choose the learning loop: representation, prediction heads, uncertainty method, acquisition function, batch diversity, and update cadence.
- Lock independent validation data and prospective success criteria before optimization.
- Deliver an executable experiment matrix, decision gates, risk register, and minimum viable loop.
Rules
- Do not optimize a proxy without testing whether it predicts the desired biological outcome.
- Keep safety, efficacy, manufacturability, and uncertainty as separate objectives until a justified decision rule combines them.
- Include baselines, ablations, blinded prospective tests, and failure analysis.
- Distinguish model improvement from therapeutic progress.
Gives 0 of the 12 instructions most plan spec skills give in 247 tokens
Counted across 1,100 of the 1,860 authors here whose files we hold, read 2026-08-06
- ask one question at a timein 46 of 1100, across 38 files
- Break plans into vertical slicesin 28 of 1100, across 10 files
- Publish issues in dependency orderin 27 of 1100, across 9 files
- Iterate until user approves the breakdownin 24 of 1100, across 6 files
- Explore the repository to understand the codebase statein 24 of 1100, across 7 files
- Use domain glossary vocabularyin 23 of 1100, across 5 files
- Apply correct triage labels to published issuesin 23 of 1100, across 5 files
- Write failing tests before implementation codein 23 of 1100, across 18 files
- Prefer AFK slices over HITLin 22 of 1100, across 7 files
- ask clarifying questions until requirements are concretein 21 of 1100, across 13 files
- Respect existing architecture decision recordsin 20 of 1100, across 5 files
- write a specification before writing any codein 20 of 1100, across 12 files
Said here and by no other author read
- read program design references
- define target profile and biological decision
- separate variables, constraints, and validation endpoints
- create a staged assay funnel
- assign each model and assay one explicit role
- specify the data contract
Grouped from the skills themselves: near-identical wordings counted once, and counted by distinct author, so one author publishing three of these counts once. Length counted with cl100k_base; the agent that loads this file may tokenize it differently.