Repeatability
High
The structure is identical every run: ingest PDFs from known folders, extract the same fixed fields, deduplicate, output CSV. This is a textbook batch-processing pipeline with no instance-level judgment variation.
Ambiguity Tolerance
High
Success criteria are concrete — six named fields, one CSV, flagged failures. The deduplication logic (likely email or name+phone match) needs a one-time definition, but once set, the agent knows exactly when it's done.
Data & Tool Availability
High
The agent needs read access to three local or cloud folders and write access to output a CSV — both are straightforward to provision. PDF parsing libraries (pdfplumber, PyMuPDF) and regex/NLP extraction are mature and readily available.
Error Cost
Low
The output is a CSV reviewed by a human before ATS import, so errors are catchable before they cause downstream harm. Worst case is a missed candidate or a wrong field value, both correctable with a spot-check.
Human Judgment Required
Low
Extracting structured fields from resumes is a solved NLP problem. The only edge cases — ambiguous deduplication matches or malformed PDFs — are handled by flagging rather than requiring real-time human judgment.