thyrox

Start with the DISCOVER stage to explore context, stakeholders, and symptoms before planning. THYROX then guides work through analysis, strategy, scoping, decomposition, and execution stages, each producing artifacts that feed the next.

Use the PLAN EXECUTION stage (decompose) to create atomic tasks with IDs like [T-NNN], each traced back to a requirement (R-N). The task plan includes a dependency graph and lives in the work package directory.

Stage count scales with work size: micro fixes use stages 1, 10, 11; small features add stage 3; medium work adds strategy and decomposition; only large multi-session projects run all 12 stages.

Use DMAIC for reducing defects with statistical data, Lean for eliminating visible waste like time and inventory, and PDCA for fast iterative improvement cycles. The methodology-selection-guide reference provides decision trees for similar candidates.

Artifacts live in timestamped directories under context/work/YYYY-MM-DD-HH-MM-SS-name/, with flat phase-based subfolders (discover/, analyze/, plan/, execute/, track/) containing markdown files with YAML metadata blocks.

Combining run_in_background with output redirection empties the task output file, causing manual polling loops to run forever as zombie processes. Use notification-driven completion or synchronous execution with a hard timeout instead.