hydromind-research-execution
Stage 2 main skill of the HydroMind Research family. Guide one session through batch-first autonomous research using a file-backed `research/<research_id>/` workspace. `context.json` is the canonical short-context source, `study_design.json` (when present) is the planning anchor that biases batch selection toward declared evidence gaps, `hydro queue wait-idle` is the synchronization boundary, `feedback.md` is a lightweight friction evidence log, and `hydro research paper status` signals when to hand off to Stage 3. Use this whenever the researcher wants to "continue the research", "run the next batch", "close an evidence gap", "resume the research workspace", "plan the next round of experiments", or otherwise advance evidence collection - even without the word "execution". Use this AFTER `hydromind-research-design` has written `study_design.json` with drafting-blocking evidence gaps. When `hydro research paper status` reports `drafting_ready`, hand off to `hydromind-research-paper`.
Repository SourceNeeds Review
hydromind-research-explore
Stage 1 main skill of the HydroMind Research family. Enter integrated study exploration - a thinking partner that helps the researcher firm up "what experiments to run AND what paper to tell" together, grounded in the existing `research/<research_id>/` workspace (`context.json`, `draft.md`, optional existing `study_design.json`, optional `paper/inputs/experimental_log.md`). Use this whenever the researcher is unsure what the next paper should claim, which evidence gaps still block drafting, which literature clusters must be covered, what figures the paper needs, or how Methods/Experiments/Discussion should be framed - even if they don't explicitly say "design" or "explore". Use this BEFORE `hydromind-research-design` materializes `study_design.json`, and BEFORE `hydromind-research-execution` commits to a new batch of runs. This skill MUST stay read-only on canonical state.
Repository SourceNeeds Review
hydromind-research-paper
Stage 3 main skill of the HydroMind Research family. Drive a research paper from a validated `study_design.json` through the canonical `paper/` subworkspace - projecting `paper/inputs/idea.md` and `paper/outline.json` from the design, building the verified literature pool, preparing figures, producing a constrained unified draft, and running accept/revert/halt refinement to a `final/paper.md`. Use this whenever the researcher wants to "start the paper", "do literature for the paper", "draft the paper", "refine the paper", "finalize the manuscript", or otherwise advance any phase of the paper pipeline - even without the word "workflow". Use this AFTER `hydromind-research-design` has written and validated `study_design.json` with no drafting-blocking evidence gaps (otherwise route to `hydromind-research-execution` first). Inside drafting (Phase 4) and refinement (Phase 5), defer prose-quality decisions to the internal sub-skill `hydromind-research-paper-writing`; deterministic gates (literature verification, ci
Repository SourceNeeds Review
hydromind-research-paper-writing
Stage 3 internal prose-quality sub-skill of the HydroMind Research family. Improve prose quality of an ML/CV/NLP-style paper draft — section structure, paragraph flow, claim-evidence alignment, and reviewer-facing presentation. Loaded by `hydromind-research-paper` Phase 4 (unified drafting) and Phase 5 (refinement) to handle the writing-craft layer. MUST NOT be invoked standalone for HydroMind research-to-paper work, because deterministic gates (literature verification, citation coverage, anti-leakage, refinement decision/halt) live in `hydromind-research-paper` and must not be bypassed. If a researcher asks for paper help without going through `hydromind-research-paper` first, route through the parent workflow.
Repository SourceNeeds Review