merge-help-csv.py: fallback HEADER uses after/before, all module-help.csv files use preceded-by/followed-by

[stefanoginella]

Describe the bug

merge-help-csv.py declares its fallback HEADER with after / before in positions 9–10, but every module-help.csv in this repo uses preceded-by / followed-by. When the fallback is used, the written CSV has two misnamed columns, and any consumer reading preceded-by / followed-by finds nothing.

It's a clean internal inconsistency — verified against main @ e6935f2, all five copies of the script disagree with all five CSV assets:

merge-help-csv.py (all say after / before)	module-help.csv (all say preceded-by / followed-by)
skills/bmad-module-builder/assets/standalone-module-template/	skills/module-help.csv
skills/bmad-module-builder/assets/setup-skill-template/scripts/	skills/bmad-module-builder/assets/setup-skill-template/assets/
skills/bmad-bmb-setup/scripts/	skills/bmad-bmb-setup/assets/
samples/sample-module-setup/scripts/	samples/sample-module-setup/assets/
samples/bmad-agent-dream-weaver/scripts/	samples/bmad-agent-dream-weaver/assets/

Because the two template copies are what scaffolded modules vendor, downstream modules inherit it — that's how it was found.

Steps to reproduce

The fallback is:

header = target_header if target_header else (source_header if source_header else HEADER)

Reaching it is harder than it looks, which is why this is latent rather than user-visible. read_csv_rows returns rows[0] unconditionally, so a falsy source_header normally implies an empty or missing source — which trips the sys.exit(1) just above. The one path through is a source whose first line is blank (csv.reader yields [], falsy) plus a non-existent target:

printf '\nabm,skill-a,A,1,desc,x,y,z,,,,,\n' > /tmp/src.csv
python3 merge-help-csv.py --target /tmp/fresh.csv --source /tmp/src.csv
head -1 /tmp/fresh.csv
# module,skill,display-name,menu-code,description,action,args,phase,after,before,required,output-location,outputs
#                                                                   ^^^^^ ^^^^^^

A normal merge never hits it, since either file supplies a header. So the practical cost isn't a broken install today — it's a wrong schema constant sitting in five files, which the next person to reach for HEADER (to add column validation, say) inherits.

Expected behavior

HEADER matches the schema the rest of the toolchain emits — preceded-by / followed-by in positions 9–10.

Environment

• bmad-builder main @ e6935f2776c34ffd9b18f4ef1e334584ef10a50f (npm bmad-builder@1.1.0)

PR

Happy to open a one-line-per-file PR if that's welcome.

Additional context

Credit to @hugheba, who found this while vendoring a scaffolded module downstream and reported it at stefanoginella/auto-bmad#8 — verified by reading installed files, not static analysis. A copy of the same script in bmad-code-org/bmad-loop carries it too.

[coderabbitai]

🔗 Related PRs

#36 - fix: complete setup skill with legacy cleanup, result templates, and path fixes [merged]
#38 - feat: add module-help.csv and module.yaml to src/skills [merged]
#92 - fix: align validator and docs with preceded-by/followed-by help-CSV columns [merged]
#94 - fix(setup): resolve {project-root} in script path arguments [merged]
#97 - fix(validate-module): harden standalone module validation [merged]

---

📝 Issue Planner

_{Check the box below or use the @coderabbitai plan command to generate an implementation plan and prompts that you can use with your favorite coding assistant.}

• Create Plan

---

🧪 Issue enrichment is currently in open beta.

You can configure auto-planning by selecting labels in the issue_enrichment configuration.

To disable automatic issue enrichment, add the following to your .coderabbit.yaml:

issue_enrichment:
  auto_enrich:
    enabled: false

💬 Have feedback or questions? Drop into our discord!