Fallback workflow for extracting text from PDFs when read_file returns binary data
72%
Does it follow best practices?
Run evals on this skill
Adds up to 20 points to the overall score
Low
Low-risk findings worth noting
Fix and improve this skill with Tessl
tessl review fix ./benchmarks/gdpval/skills/pdf-text-extraction-fallback-85d5ca/SKILL.mdUse this skill when read_file returns binary data or garbled content for PDF files instead of readable text. This workflow provides a reliable fallback using command-line PDF tools.
read_file with filetype: pdf returns binary data, unreadable characters, or errorsAfter attempting to read a PDF with read_file, check if the output is:
# Example of problematic output from read_file
%PDF-1.4
1 0 obj
<< /Type /Catalog ...If the output looks like raw PDF structure or binary, proceed to Step 2.
Invoke shell_agent to extract text using pdftotext (preferred) or pdfplumber (Python fallback):
Task: Extract all text content from <filename.pdf> using pdftotext or pdfplumber.
Output the extracted text in readable format. If pdftotext is not available, use Python with pdfplumber library.Example shell_agent invocation:
shell_agent task="Extract text from Move_Out_Inspection_Tracker.pdf using pdftotext. Save output to a .txt file and return the content."After extraction, validate that the content contains expected text patterns:
# Validation checklist
def validate_pdf_extraction(text, expected_patterns=None):
checks = [
bool(text.strip()), # Not empty
len(text) > 50, # Has substantial content
not text.startswith('%PDF'), # Not raw PDF structure
]
if expected_patterns:
for pattern in expected_patterns:
checks.append(pattern.lower() in text.lower())
return all(checks)Common expected patterns to check:
If validation fails:
Try alternative tool: If pdftotext failed, try pdfplumber:
shell_agent task="Extract text from <file.pdf> using Python pdfplumber library. Handle any encoding issues."Try OCR fallback: For scanned PDFs:
shell_agent task="This PDF may be scanned. Use pytesseract or similar OCR tool to extract text from <file.pdf>."Report specific error: Document what patterns were expected but not found.
Once validated text is obtained:
# Complete extraction workflow
def extract_pdf_text_fallback(pdf_path, expected_patterns=None):
"""Extract text from PDF with fallback handling."""
# Step 1: Try read_file first
content = read_file(filetype="pdf", file_path=pdf_path)
# Step 2: Check if binary/unreadable
if is_binary_or_garbled(content):
# Step 3: Use shell_agent fallback
result = shell_agent(
task=f"Extract all text from {pdf_path} using pdftotext. Return the text content."
)
content = result.stdout
# Step 4: Validate
if not validate_pdf_extraction(content, expected_patterns):
# Try pdfplumber as secondary fallback
result = shell_agent(
task=f"Extract text from {pdf_path} using Python pdfplumber library."
)
content = result.stdout
return content
def is_binary_or_garbled(text):
"""Check if text appears to be binary or unreadable."""
if not text:
return True
if text.startswith('%PDF'):
return True
# Check for high ratio of non-printable characters
non_printable = sum(1 for c in text if ord(c) > 127 or ord(c) < 32)
return non_printable / len(text) > 0.3HKUDS/OpenSpace
3827781
If you maintain this skill, you can claim it as your own. Once claimed, you can manage eval scenarios, bundle related skills, attach documentation or rules, and ensure cross-agent compatibility.