| name | scan-report |
| description | Use when user wants to Summarize the last dry-run scan output from /tmp/scan-diag.txt and classify failures by type |
Read and classify the most recent dry-run output:
cat /tmp/scan-diag.txt 2>/dev/null || echo "No scan-diag.txt found - run: uv run autoscrapper scan --dry-run 2>&1 | tee /tmp/scan-diag.txt"
Classify each outcome line:
SKIP_UNLISTED, Meaning=OCR read a name but no rule matched, Fix path=Check rule store, fuzzy threshold, item name
SKIP_EMPTY, Meaning=Slot appeared empty, Fix path=Expected - verify count is reasonable for grid size
UNAVAILABLE, Meaning=Context-menu OCR matched "Unavailable" button, Fix path=Check startswith("unavailable") guard in ocr_context_menu
Action taken (SELL/KEEP/RECYCLE), Meaning=Rule matched and action dispatched, Fix path=Verify correct rule won (custom vs default precedence)
infobox not found / timeout, Meaning=Page detection failed, Fix path=Route to scan-validator agent
Garbled item name, Meaning=OCR misread, Fix path=Route to ocr-reviewer agent
Output a structured report:
- Outcome table - counts per outcome type
- Top 3 issues - most actionable failures with item name + outcome
- Suggested fix path - which skill or agent to invoke next
Related skills: /diagnose-scan, /ocr-debug, /scan-failed