algorithm-prototyping-workflow

Installed Root

Resolve the installed reasflow-dev skills root before running packaged scripts:

REASFLOW_SKILLS_ROOT="${REASFLOW_SKILLS_ROOT:-}"
if [ -z "$REASFLOW_SKILLS_ROOT" ]; then
  if [ -d ./.agents/skills ]; then
    REASFLOW_SKILLS_ROOT="$(pwd)/.agents/skills"
  elif [ -d "$HOME/.agents/skills" ]; then
    REASFLOW_SKILLS_ROOT="$HOME/.agents/skills"
  else
    echo "reasflow-dev skills not found in ./.agents/skills or $HOME/.agents/skills" >&2
    exit 1
  fi
fi

Algorithm Prototyping Workflow

Overview

Use this when Algorithm must deliver concrete method artifacts, not only discussion. The minimum bar is pseudocode plus runnable code plus one executed feasibility run.

Required Artifacts

1. Alg_Exp/document/pseudocode.tex
2. Alg_Exp/code/algorithm.py
3. Alg_Exp/code/test_algorithm.py
4. Alg_Exp/code/quick_experiment.py (or equivalent experiment section in test code)
5. Alg_Exp/data/test_run.json or Alg_Exp/data/test_run.csv
6. Alg_Exp/document/algorithm_design.md
7. Alg_Exp/document/test_experiment_report.md

Workflow

1. Lock assumptions and invariants first with algorithm-design-review.
2. Draft and save pseudocode before implementation.
3. Implement the minimal runnable algorithm API.
4. Add and run at least one basic executable test.
5. Run one quick experiment on a controllable toy problem and capture logs.
6. If numeric traces are available, use smart-plotting to save at least one figure under Alg_Exp/picture/.
7. Record commands, metrics, pass or fail judgment, and next steps in test_experiment_report.md.

Command Logging

Use experiment-execution for reproducible command logging and seed tracking whenever scripts are run from Alg_Exp/code/.

Final Response Contract

Before finishing, include:

• a short pseudocode snippet
• the executed test command and result
• the executed quick experiment command and key metrics
• an explicit feasibility judgment
• paths to all generated artifacts