FABRICATION_WATCHLIST substring match causes false positives (e.g. "rust" in "TrustSec")

[tearl42]

Description of the bug:
The fabrication detection in validator.py uses a simple in substring check against the skills text, which causes false positives when a watchlist term appears as part of a longer word.

To Reproduce:
Add TrustSec to your skills boundary. Run the tailor stage. Every attempt will fail with Fabricated skill: 'rust' even though "rust" was never added — it's being matched inside "TrustSec".

Expected behavior:
The check should use word boundary matching so "rust" only matches the standalone word "rust", not substrings within other words.

Fix
In validator.py around line 132, change:
if fake in skills_text:
to
if re.search(r"\b" + re.escape(fake) + r"\b", skills_text):

Environment:
ApplyPilot version: (your version)
Model: gemini-2.5-flash
Resume: Long-career networking professional with TrustSec in skills boundary