Conflict checker skill

This skill wraps the duplicate_checker.py script with additional business logic for submission approval.

Approval flow

When a submission request arrives from Jay or Rick (via EM):

1. •Parse the request: consultant_id, end_client, vendor_name, job_posting_id
2. •Load the consultant's do-not-submit list from their profile
3. •Load the master DNS list from memory/do-not-submit-master.md
4. •Load full submission history from memory/submission-log.jsonl
5. •Run duplicate_checker.py with all inputs
6. •If BLOCK: respond with NO + reason + conflicting record reference
7. •If WARN: respond with advisory warning + allow human to decide
8. •If ALLOW: respond with YES + generate submission record ID

Additional checks beyond duplicate_checker.py

• •LinkedIn profile daily limit: check if the assigned profile has exceeded its daily application count
• •LinkedIn profile health: check if the profile is GREEN (from Z's tracking data)
• •Consultant application limit: check if the consultant has already hit 4-5 applications today
• •Rate compatibility sanity check: flag if posted rate is more than 20% below consultant minimum

Response format

SUBMISSION APPROVAL RESPONSE
=============================
Decision: APPROVED / BLOCKED / WARNING
Submission ID: [auto-generated if approved]
Consultant: [name] (ID)
Job: [title] (posting ID)
Client: [end_client] via [vendor]

Checks passed: [list]
Checks failed: [list with reasons]
Conflicts: [list with record references]