Phase guidance for the neuroflow /pipeline command. Loaded automatically when /pipeline is invoked to orient agent behavior for planning and executing multi-step research pipelines.
From neuroflownpx claudepluginhub stanislavjiricek/neuroflow --plugin neuroflowThis skill uses the workspace's default tool permissions.
Dispatches parallel agents to independently tackle 2+ tasks like separate test failures or subsystems without shared state or dependencies.
Executes pre-written implementation plans: critically reviews, follows bite-sized steps exactly, runs verifications, tracks progress with checkpoints, uses git worktrees, stops on blockers.
Guides idea refinement into designs: explores context, asks questions one-by-one, proposes approaches, presents sections for approval, writes/review specs before coding.
The /pipeline command orchestrates a sequence of neuroflow commands in order — either interactively (pause and confirm at each step) or in brutal mode (--executor, run straight through without stops).
flow.md for completed phases and project_config.md for the active phase and tools.[pending] and let the user decide.pipeline-plan.md, read it first, show the current status, and continue from where execution stopped.| Behaviour | Interactive | Brutal (--executor) |
|---|---|---|
| Pause between steps | ✅ Always | ❌ Never |
| Ask where to stop | ✅ Yes (Step 3b) | ❌ No — all pending steps run |
| User can skip individual steps | ✅ Yes | ❌ No — all pending steps run |
| User can stop mid-pipeline | ✅ Yes | ❌ Errors logged; pipeline continues |
| Error handling | Stop and ask | Log and continue; report at end |
| Summary at end | ✅ Yes | ✅ Yes (more detailed) |
After the plan is confirmed (Step 3b in the command), present numbered choices for each pending phase plus an "all the way through" option. The user's answer determines the stop_after value:
[deferred] in pipeline-plan.md. Save Stop after: in the plan header. Execute only up to the stop point.When resuming, [deferred] steps are surfaced prominently so the user knows there is more to run. Re-run Step 3b on resume to let the user pick a new (or extended) stop point.
When the user selects the full pipeline covering the complete research journey (ideation → paper), generate a fresh, original joke on the spot. Rules:
Use this as the default order when inferring a pipeline from project state:
ideation → grant-proposal (optional) → experiment (optional) → tool-build (optional) → tool-validate (optional) → data → data-preprocess → data-analyze → paper
Brain simulation phases insert between data-analyze and paper:
… → data-analyze → brain-build → brain-optimize → brain-run → paper → …
Phases marked optional should only be included if:
.neuroflow/ or the repoThe pipeline-plan.md file is the canonical record of what was planned and what happened. Keep it up to date as execution proceeds — treat it as a live log, not a one-time snapshot.
# Pipeline plan
Generated: YYYY-MM-DD
Mode: interactive | brutal (--executor)
Stop after: /data-analyze | all the way through
## Steps
| # | Command | Status | Notes |
|---|---|---|---|
| 1 | /ideation | done | Completed YYYY-MM-DD |
| 2 | /data | pending | — |
| 3 | /data-analyze | pending | — |
| 4 | /paper | deferred | Beyond stop point — resume to continue |
Valid status values: pending, done, skipped, error, deferred
neuroflow:neuroflow-core — read first; defines the command lifecycle, .neuroflow/ write rules, and the end-of-command checklistneuroflow:neuroflow-develop — relevant when the pipeline includes plugin development steps or when the user is building something inside neuroflow itselfreasoning/pipeline.json at the start and update it at the end — this pipeline definition is a significant project-level decisionreasoning/pipeline.json — it is a deliberate scope decisionpipeline-plan.md after every step, not just at the endreasoning/pipeline.json entry recording the change and why.neuroflow/pipeline/