an-pipeline-runner

name: an-pipeline-runner description: Runs the full Lb2Lgamma BR analysis pipeline autonomously step-by-step - from environment setup through preselection, MC corrections, BDT, gPID, mass fit, efficiency, and branching-ratio extraction. Use when the user asks to run the pipeline, execute analysis steps, or resume the analysis from a specific point.

Analysis Pipeline Runner

Autonomous agent for executing the Lb->Lambda gamma branching-ratio measurement pipeline. Each step is self-contained: verify prerequisites, run, check outputs, report. The agent stops on first failure and proposes a fix.

Pipeline overview

Step 0  Environment setup        source setLCG.sh
Step 1  Preselection             python/preselection/
Step 2  MC corrections           python/mc_corrections/ (Snakemake)
Step 3  BDT training + apply     python/bdt/
Step 4  Photon PID (gPID)        python/preselection/ (cut file)
Step 5  Mass fit                 python/massfit/
Step 6  Efficiency               python/massfit/calculate_efficiencies.py
Step 7  Branching ratio          python/massfit/calculate_br_magdown.py
Step 8  Validation               compare to reference values

Guiding principles

1. Run in small verified steps. After each command, check exit code and expected output before proceeding.
2. Stop on first failure. Print the failing command, its stderr, and a proposed fix. Do not continue until the user confirms.
3. Never overwrite data silently. All writes go to validation_out/ or documented output paths. Ask before overwriting existing files.
4. Confirm before long-running steps. BDT training and Snakemake can take hours. Show the command and estimated wall time; wait for YES.
5. Use per-step skill files for detailed checklists. Call get_skill(skill="an-pipeline-runner", step="<name>") before each step for the exact commands and validation criteria.
6. Use the orchestrator script for real runs. Prefer scripts/run_step_with_corrector.sh for pre-flight checks, checkpointing, feedback capture, and rollback.

Progress checklist

Lb->Lambda gamma Pipeline Progress
- [ ] Step 0: Environment setup
- [ ] Step 1: Preselection
- [ ] Step 2: MC corrections
- [ ] Step 3: BDT training + application
- [ ] Step 4: gPID cut
- [ ] Step 5: Mass fit
- [ ] Step 6: Efficiency calculation
- [ ] Step 7: Branching-ratio extraction
- [ ] Step 8: Validation against reference

Per-step details