normalize-apps

name	normalize-apps
description	Normalize existing application folders to the canonical output naming standard - rename leading-date slugs, move flat files into sub-folders, fix drifted filenames, fold stray latex outputs, and backfill output_type front matter. Dry-run preview with user-editable mapping.
disable-model-invocation	true
argument-hint	[--dry-run]

JobOps Application Folder Normalization

This skill brings already-migrated v2.0 application folders into line with the canonical output naming standard ({Company}_{Role}_{YYYYMMDD} slugs; fixed sub-folder filenames; output_type front matter on every Markdown output). It is distinct from /jobops:migrate, which moves legacy v1.x top-level folders (OutputResumes/, Briefing_Notes/, …) into the v2.0 layout — run that first if those exist.

Use this skill when application folders already live under applications_root but drifted in shape: leading-date slug names, improvised filenames (resume_step3_final.md, Step3_Final_Resume_*.md), flat root files that were never moved into sub-folders, a stray latex/ folder, or Markdown outputs missing their output_type key. After it runs, /jobops:dashboard should show zero false ○ for work that actually exists.

Flags:

• --dry-run — stop after the preview without renaming or moving any files.

---

Configuration

Read .jobops/config.json. If missing, stop with:

JOBOPS NOT CONFIGURED Run /jobops:setup to initialize your workspace.

Use config.directories.applications_root as the scan root and config.directories.company_intelligence to validate company tokens. Use config.directories.<key> for all other paths.

---

Step 1: Scan application folders

List the immediate sub-directories of config.directories.applications_root whose name is not dot-prefixed. Each is one application folder. If there are none, print:

Nothing to normalize. No application folders found.

…and exit with success.

For each application folder, plan four classes of change (Steps 2–5). Collect every planned action without executing it yet.

---

Step 2: Slug / folder rename (leading-date → canonical)

For each folder name, decide the canonical slug {Company}_{Role}_{YYYYMMDD}:

• Already canonical — leading PascalCase company token, trailing 8-digit date, no leading date/time prefix → no rename.
• Leading date+time prefix ^(\d{4})-(\d{2})-(\d{2})_(\d{6})_(.+)$ (e.g. 2026-04-15_214414_GTAA_DirectorAirportCommercialRealEstate) → move the date to the tail as a compact 8-digit date and drop the time: GTAA_DirectorAirportCommercialRealEstate_20260415.
• Leading date only ^(\d{4})-(\d{2})-(\d{2})_(.+)$ → same, dropping no time component.
• Validate the resulting leading token against company_intelligence sub-folders (slugified, case-insensitive). If it does not match any company folder, keep the rename but flag it in the preview as company-token-unverified so the user can correct it via the editable mapping.

A folder whose target slug collides with an existing folder is added to the unresolved list (never silently merged).

---

Step 3: Move flat root files into sub-folders + rename to canonical

Within each application folder, map root-level (and otherwise mis-placed) files to their canonical destination by filename pattern (case-insensitive; tried in order):

Source pattern (anywhere in the app folder)	Canonical destination
Assessment_*.md, assessment_report*.md	assessment/assessment.md
Rubric_*.md, *_rubric.md	assessment/rubric.md
*domain_research*.md	assessment/domain_research.md
Step1_Draft_*.md, resume_step1*.md, *step1*draft*.md, *_draft.md	resume/step1_draft.md
Step2_Provenance*.md, resume_step2*.md, resume_provenance.md, *provenance*.md	resume/step2_provenance.md
Step3_Final_Resume_*.{md,pdf,tex,docx}, resume_step3_final.{md,…}, resume_final.{md,…}, *step3*final*.{md,…}, *_final.{md,…} (excluding *provenance*)	resume/step3_final.{md,pdf,tex,docx}
Step4_CoverLetter_*.md, CoverLetter_*.md, cover_letter_*.md, *cover*letter*.{md,pdf,tex,docx}	cover-letter/cover_letter.{md,…}
Briefing_*_Part{N}_*.md, briefing_part{N}*.md	interview/briefing_part{N}.md
Briefing_*.md (no part marker)	interview/briefing.md
InterviewPrep_*.md, interview_prep_*.md	interview/interview_prep.md

Notes:

• A PDF/TeX/DOCX derivative shares the exact basename of its canonical .md source and lands in the same sub-folder (resume/step3_final.pdf, cover-letter/cover_letter.docx).
• job_posting.md at the app root stays put — only its front matter is touched (Step 6).
• If two source files would map to the same canonical destination (e.g. two *final*.md), add both to the unresolved list rather than overwriting.

---

Step 4: Rename drifted structured filenames

For files already inside the correct sub-folder but under a non-canonical name, rename to the canonical basename (resume/resume_step3_final.md → resume/step3_final.md, resume/resume_final.md → resume/step3_final.md, cover-letter/cover.md → cover-letter/cover_letter.md, etc.). The Step 3 patterns also catch these; treat Step 4 as the in-sub-folder pass so nothing is missed.

---

Step 5: Fold stray latex/ outputs

If an app folder contains a latex/ sub-folder, move each output into the matching content sub-folder, basename-matched: a resume PDF/TeX → resume/, a cover-letter PDF/TeX → cover-letter/, named to match the canonical .md basename in that sub-folder (latex/resume.pdf → resume/step3_final.pdf). Remove the now-empty latex/ folder. If the target sub-folder has no canonical .md to match against, add the file to the unresolved list.

---

Step 6: Backfill output_type front matter

For each Markdown file now at a canonical location, ensure its YAML front matter carries the correct output_type (the durable, filename-independent detection key). If front matter is absent, wrap a new block; if present but missing the key, add it. Infer from location:

Canonical location	output_type
job_posting.md	job_posting
assessment/rubric.md	rubric
assessment/assessment.md	assessment
resume/step1_draft.md	resume_step1
resume/step2_provenance.md	resume_provenance
resume/step3_final.md	resume_final
cover-letter/cover_letter.md	cover_letter
interview/briefing*.md	briefing
interview/interview_prep*.md	interview_prep

Never overwrite an existing output_type value — only add it where missing.

---

Step 7: Preview

Print the full plan grouped per application folder, in this order: slug rename (Step 2), file moves/renames (Steps 3–5), and output_type backfills (Step 6). Then print the unresolved list (collisions, unverified company tokens, ambiguous matches). Include total counts of each action class.

If --dry-run was passed, stop here and exit with success.

---

Step 8: User-editable mapping

Offer:

1. Execute as shown — run every planned action.
2. Edit mapping — drop into an inline editor where the user can reassign a specific file to a different destination, correct a company token, or mark any action as skip.
3. Cancel — exit without changing anything.

---

Step 9: Execute (reversible, fail-soft)

Process slug renames (Step 2) before in-folder moves so destination paths are stable. For each confirmed action:

• mkdir -p <destination parent> if needed.
• If the path is tracked in git (git ls-files --error-unmatch <path> returns 0), use git mv <src> <dst>; otherwise plain mv.
• For Step 6 backfills, edit the file in place (front-matter insert/merge only — never touch the body).
• Count failures separately and report them; do not abort the whole batch on one failure.

Reversibility: because every action is a git mv / in-place edit on a tracked tree (or a plain mv otherwise), the user can revert with git checkout/git restore. State this in the summary.

---

Step 10: Summary

Print:

• Folders scanned; slugs renamed; files moved/renamed; latex/ folders folded; output_type keys backfilled.
• Actions succeeded vs failed.
• The still-unresolved list (for manual handling).
• Reminder to run /jobops:dashboard to confirm zero false ○, and to review and commit the changes if the workspace is under version control.

Exit.