Skill falls back to system Python when .graphify_python is missing

[todd-w-shaffer]

Bug

When graphify is installed via pipx, the skill's Step 1 writes the correct Python interpreter path to graphify-out/.graphify_python. All subsequent steps use $(cat graphify-out/.graphify_python) to run against the pipx venv where networkx and other dependencies live.

However, Step 9 cleans up .graphify_python along with the other temp files. This means any follow-up invocation (--update, --cluster-only, query, path, explain) that doesn't re-run Step 1 will fall back to $(which python3) — the system Python — which doesn't have networkx installed:

ModuleNotFoundError: No module named 'networkx'

Repro

1. Install via pipx install graphifyy
2. Run /graphify . (works fine — Step 1 resolves the pipx interpreter)
3. Run /graphify query "something" (fails — .graphify_python was cleaned up, falls back to system Python)

Suggested fix

Every skill entry point (not just the full pipeline) should resolve the correct interpreter at the top. Either:

1. Don't clean up .graphify_python in Step 9 — it's tiny and useful across invocations
2. Or add the interpreter detection snippet to the start of every subcommand (query, path, explain, --update, --cluster-only)

Option 1 is simpler and covers all cases.

Environment

• macOS (Homebrew Python 3.14, externally managed)
• graphify installed via pipx install graphifyy (v0.3.15)
• Claude Code CLI

[safishamsi]

Fixed in v0.3.16 — .graphify_python is no longer deleted in Step 9 cleanup across all 6 skill files. Follow-up commands (query, --update, --cluster-only) now correctly reuse the pipx interpreter from the initial run.