phase-tool-build

The tool-build phase designs and implements lab tools or software pipelines — real-time systems, data acquisition, LSL integrations, or custom analysis pipelines.

Approach

• Define done-criteria and the tool specification before writing any code
• Ask which type of tool is needed (acquisition, real-time processing, LSL, analysis pipeline, other)
• Write tool-spec.md first; get confirmation before implementation begins
• Prefer tested, modular components over monolithic scripts
• For real-time and closed-loop tools, make the latency budget a measurable done-criterion and stamp outputs with their source timestamp (/tool-build → Real-time and closed-loop tools)
• Stimulation hardware is started by a person; no software check is a safety interlock

Relevant skills

• neuroflow:neuroflow-core — read first; defines the command lifecycle and .neuroflow/ write rules
• neuroflow:pupil-labs-neon-realtime — if the tool involves Pupil Labs Neon hardware, use this skill for device discovery, multi-threaded streaming architecture, and real-time data collection
• neuroflow:phase-tool-validate — its scripts/lsl_check.py lists LSL streams and measures rate, gaps and latency against a budget

Workflow hints

• All code goes to output_path (tools/), not inside .neuroflow/
• Save tool-spec.md to .neuroflow/tool-build/ before writing any implementation
• Note key technical decisions (library choice, architecture, interfaces) in .neuroflow/reasoning/tool-build.jsonl

Slash command

/neuroflow:tool-build — runs this workflow as a slash command.