Research

Use this skill for a requested prior-work brief or research program. Ordinary implementation does not require a research workflow.

Choose the record

Return a compact brief in the conversation for bounded work. For durable or cross-session work, use the task's existing issue, PR, W&B run or report, published report, or durable session channel. Do not create standalone repository design, research, plan, logbook, or snapshot files. If the user explicitly requests a particular artifact, honor that request.

Update docs/ or the relevant OPS.md when the result changes reusable product or operational guidance. Keep raw logs and dense data in their source systems.

Investigate

1. State the question, decision, and stopping condition.
2. Search the current checkout for task-local context. Search Echo for Marin context unless the same logical session already completed a relevant search and the question, repository scope, and freshness requirements have not changed. Reuse and cite those results instead of repeating the search. Use primary external sources when the question benefits from outside evidence.
3. Test the leading explanation against contradictions, negative results, and materially different operating regimes.
4. Stop when additional sources or experiments no longer change the decision, or when the requested effort is exhausted.

For experiment results, record the exact command, source revision, material configuration, hardware, baseline, result, and interpretation. Label estimates and exploratory results. Record failures that rule out a hypothesis; omit routine debugging history.

Use W&B for scalar series, plots, large comparison tables, or raw artifacts that are too dense for the narrative record. Runs that require direct comparison must share a project. Verify row counts, key uniqueness, aggregation, and the numbers cited in the narrative before publishing a claim.

Report

Scale the response to the decision. Include:

• the question and conclusion;
• evidence and source links;
• contradictions, limitations, and confidence;
• the next experiment only when it could change the decision.

Do not require a special branch, experiment ID, hypothesis queue, update cadence, issue, tag, or W&B project unless the task itself needs one.