fix(skill-creator): support direct execution of package_skill.py and update usage paths - #1681

Hi @maheshmurag, whenever you have a moment, could you please take a look at this quick fix? It is a small self-contained change (+13/-6) resolving standalone script execution for package_skill.py while preserving existing module imports. All verification steps have been tested and passed.

Hi @maheshmurag, just following up on this small self-contained fix. I've rebased the branch onto the latest \main\ so it is clean and ready to merge whenever you have a moment. Thank you!

Hi @Kuldeeep18 — thanks for the fix, and for the rebase: the branch now sits on the current main tip (34040c9), single commit, MERGEABLE. Reviewed at head f99c58a3fd3a0e12e9f7882e5d7d12d06634cd18 and ran a two-state battery with Python 3.11:

• ✅ Direct execution fails before, works after: from the repo root, python skills/skill-creator/scripts/package_skill.py skills/brand-guidelines → pre-PR file raises ModuleNotFoundError: No module named 'scripts' (exit 1); head packages brand-guidelines.skill (exit 0).
• ✅ Works from any cwd: also verified from skills/skill-creator/ (python scripts/package_skill.py ../brand-guidelines) and from an unrelated directory with an absolute script path — all succeed.
• ✅ No module-mode regression: python -m scripts.package_skill ../brand-guidelines works on both the pre-PR file and head (the not in sys.path guard makes the insert a no-op under -m).
• ✅ Docs & packaging: usage examples now point at scripts/package_skill.py / skills/brand-guidelines; no utils/package_skill.py references remain anywhere in the tree (grepped); the generated .skill contains brand-guidelines/SKILL.md + LICENSE.txt (identical 5281-byte artifact in every mode); py_compile clean; 100755 mode preserved; the diff is the single file, +13/−6.

Optional (non-blocking): run_eval.py, run_loop.py and improve_description.py carry the same from scripts.… imports — direct execution still raises the same ModuleNotFoundError (verified on run_eval.py). They're documented as python -m scripts.…, so nothing needs changing in this PR; just noting a shared bootstrap would cover them if direct execution is wanted repo-wide.

Hi @cj-ant, @rlancemartin — gentle follow-up on this standalone execution fix for \package_skill.py.

The branch is up to date with latest \main, clean, and has been verified by community reviewers across all execution modes without regression. Whenever you have a moment, could you please take a look for merge? Thank you!

Follow-up re-review at head 01b0047ee1.

Since our previous review at f99c58a3fd3a (comment above), the only change is a merge of main (3337550, claude-api docs): skills/skill-creator/scripts/package_skill.py is byte-identical to the previously reviewed revision (blob 07dc5208). Re-ran the two-state battery at the new head (Python 3.11):

Verified

• Parent state (f99c58a^) direct execution from its own repo root: ModuleNotFoundError: No module named 'scripts', exit 1; at the new head: exit 0, packages brand-guidelines.skill.
• Works from every cwd: repo root, skills/skill-creator/ (relative scripts/package_skill.py ../brand-guidelines), and an unrelated directory with absolute paths — all exit 0.
• No module-mode regression: python -m scripts.package_skill ../brand-guidelines exit 0.
• All four artifacts identical (sha256 586639af…), containing brand-guidelines/SKILL.md + LICENSE.txt.
• Diff re-checked at the new head: single file, +13/−6; py_compile clean; mode 100755 preserved; no utils/package_skill.py references remain; no network/subprocess patterns in the touched file.

Remaining: none blocking. The optional note from the earlier review still stands (siblings run_eval.py / run_loop.py / improve_description.py share the from scripts.… pattern and are documented as python -m scripts.…; a shared bootstrap would cover them if direct execution is ever wanted repo-wide).

Still LGTM for merge.