菜单
Ask the same conceptual or planning question to Codex, Claude Code, and Antigravity, then synthesize the independent perspectives. Use when the user asks for a council, cross-model opinions, independent viewpoints, or conceptual comparison rather than a code diff review.
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
基于 SOC 职业分类
Shared multi-agent harness state and coordination for Codex, Claude Code, and Antigravity. Use for repo resume/state, shared memory or active-task handoff, plan claims/recovery, fail-ledger checks, stale `.oms` cleanup, role/executor setup, local provider calls or write delegation, patch admission/landing, artifact provenance, and prior-session handoff.
Discovery-focused chem-bio ML guardrails for small molecules, molecular 3D and physics, proteins and variants, biomolecular interactions, peptides and antibodies, reactions and synthesis, generative design, cellular omics and phenotypic profiling, and biomedical networks. Use before writing, reviewing, evaluating, or training on RDKit/SMILES/SELFIES, QSAR/ADMET, quantum or force data, MMseqs2/MSA/protein language models, PDB/AlphaFold/DMS, docking/DTI/PPI, peptide-MHC or antibody data, reaction/retrosynthesis/yield data, molecular or protein generation, RNA/oligonucleotide/CRISPR, single-cell perturbation/Cell Painting, or biomedical knowledge graphs. Enforce domain splits, leakage checks, label semantics, baselines, metrics, calibration, applicability domain, and provenance.
PyTorch training correctness for optimizers and schedules, distributed/DDP, variable-size masked losses, checkpoint save/load, CUDA runtime settings, and equivariant model checks. Use when writing or reviewing training loops, optimizer/scheduler setup, multi-GPU code, loss reduction, or checkpoints.
Multi-agent code review. Use when the user explicitly requests independent verification, cross-agent/council review, or a release go/no-go or requested ML pre-training gate needs several independent reviewers. High-risk changes alone do not require this skill.
Hypothesis-driven loop for ML and scientific experiments. Use to turn a question into a falsifiable hypothesis, design the smallest disconfirming run, pre-register prediction/baseline/metric, interpret results, coordinate experiment claims, reproduce or compare runs, or trace an output to its run.
Slurm/HPC workflow helper. Use when working on clusters, partitions, nodes, sbatch/srun jobs, GPU/CPU allocations, queues, logs, checkpoints, or any resource-heavy command — e.g. "submit the job", "check the queue", "잡 돌려줘", "큐 확인", "GPU 몇 장", "학습 돌려줘" on a cluster; also digesting a long job log (job-digest, "로그 요약") and reconciling finished Slurm jobs into shared state (run-reconcile, sacct). Reads local cluster reference when generated.
| name | peer-ask |
| description | Ask the same conceptual or planning question to Codex, Claude Code, and Antigravity, then synthesize the independent perspectives. Use when the user asks for a council, cross-model opinions, independent viewpoints, or conceptual comparison rather than a code diff review. |
Goal: get three independent perspectives on the same question, then synthesize. Do not outsource judgment. Use local CLI tools only. Do not use MCP servers, app connectors, or plugin connector tools.
Use for conceptual questions, design tradeoffs, planning, alternatives, or "ask codex/claude/antigravity" requests that are not primarily code review.
For code diff review, use peer-review instead. For a one-provider
question, prefer agent-run.sh; it records task outcomes and routes read/write
automatically.
The user should be able to invoke this skill in natural language. Choose the smallest context that answers the question:
current implementation, this repo, current state, 현재 구현,
현재 상황, or asks whether the repo's structure/design is appropriate: use
--repo . --repo-context.current changes, diff, uncommitted, 이번 수정, 현재 변경,
or asks about work not yet committed: use --repo . --diff.--diff unless the diff itself matters.do not edit, 수정하지 말고, opinion only, or asks only
for explanation/evaluation, do not edit files; run ask and synthesize only.If optional repo context would include secrets, private paths, machine details,
or generated cluster details, omit that context and say so in Verification.
Sending the prompt and optional sanitized repo context to codex, claude, and
antigravity local CLIs is allowed by default for this skill. Do not include secrets,
private keys, local machine details, generated cluster details, or private paths.
By default no repo context is attached. Use --repo-context only when current
repo status matters, and --diff only when the diff is needed.
~/.oh-my-setting/scripts/peer-ask.sh \
--prompt "Compare RAG and fine-tuning tradeoffs for this project."
Optional repo context:
~/.oh-my-setting/scripts/peer-ask.sh \
--repo . \
--repo-context \
--prompt "Given this repo state, what are the next implementation risks?"
Optional sanitized diff:
~/.oh-my-setting/scripts/peer-ask.sh \
--repo . \
--diff \
--prompt "What design alternatives does this diff suggest?"
Artifacts are written under .oms/artifacts/ask/.
When policy forbids sending repo context to an external provider, use
--export-only; run the exported prompt where allowed, then import the answer
with import-agent-result.sh. To recover a recent run, use
artifact-index.sh latest or artifact-index.sh failures.
Synthesize compactly:
Consensus:
Divergence:
Best answer:
Caveats:
Verification:
List unavailable, blocked, or skipped providers under Verification.