Extract data from supplied spreadsheets and documents, check for missing or truncated content, and try alternative reading methods when needed. Results identify source files and clearly separate extracted facts from external information or estimates.
中文简介正在准备中。当前显示英语简介。
Ensures agents extract data from context files with validation and fallback strategies before resorting to assumptions or external searches.
安装前可以查看交给 AI 代理的指令内容。
Prevent data hallucination and inefficiency by mandating that agents inspect, validate, and fully extract data from provided reference files before attempting web searches or generating synthetic data. When extraction is incomplete, use fallback strategies before making assumptions.
If a reference file is provided in the task context, it is the source of truth. Do not fabricate data or search the web for information that may exist within the provided attachments. If extraction appears incomplete, attempt alternative methods before proceeding.
At the start of every task, explicitly list all files provided in the context window or attachment panel.
.xlsx, .csv), documents (.pdf, .docx, .pptx), or data dumps (.json, .txt, .xml, .yaml).Determine if any provided file contains the data required to complete the task.
If relevant files are found:
read_file, pandas, pdf_reader).When initial extraction is incomplete or critical data is missing, attempt these strategies in order before making assumptions:
.docx: Try python-docx directly via execute_code_sandbox if read_file was truncated..xlsx/.csv: Try pandas with explicit sheet selection or openpyxl directly..pdf: Try pdfplumber, PyPDF2, or shell-based pdftotext if available.cat, xxd, strings).shell_agent or run_shell for format-specific tools:
unzip -p file.docx word/document.xml | xmllint --format - (extract raw docx XML)in2csv file.xlsx (convert Excel to CSV via shell)pdftotext file.pdf - (extract PDF text via command line)read_file.read_file on Pricing_email.docx (truncated at 980 chars). Attempted python-docx extraction via sandbox—pricing table in section 2 not present in file. Pricing numbers unavailable from provided context."When presenting data in the final output:
Massabama_active_listings.xlsx..."pdftotext from report.pdf..."If all extraction strategies fail to provide the specific data needed:
Task: Create a report on active property listings.
Context: Massabama_active_listings.xlsx is attached.
Incorrect Approach:
Correct Approach:
Massabama_active_listings.xlsx in context.read_file or pandas.Massabama_active_listings.xlsx, the total value is..."Task: Extract pricing tiers from Pricing_email.docx.
Context: Pricing_email.docx attached, but read_file output cuts off mid-sentence.
Incorrect Approach:
Correct Approach:
read_file output ends at 980 chars, mid-sentence.execute_code_sandbox with python-docx to read full document.read_file (truncated) and python-docx (table structure not preserved)."| Tool | Known Limitations | Fallback Strategy |
|---|---|---|
read_file | May truncate at ~1000-5000 chars depending on format | Use execute_code_sandbox with format-specific library |
execute_code_sandbox | May have missing dependencies or sandbox errors | Use shell_agent or run_shell for CLI tools |
shell_agent | Slower, but more flexible with system tools | Use for pdftotext, unzip XML extraction, in2csv, etc. |
If all extraction attempts fail:
*** End Files *** Add File: examples/extraction_fallback.sh #!/bin/bash
extract_docx_raw() { local file="$1" # Extract raw XML from docx (docx is a zip archive) unzip -p "$file" word/document.xml 2>/dev/null | xmllint --format - 2>/dev/null }
extract_xlsx_to_csv() { local file="$1" # Convert Excel to CSV using in2csv (from csvkit) if command -v in2csv &>/dev/null; then in2csv "$file" 2>/dev/null else echo "in2csv not available; try python pandas approach" return 1 fi }
extract_pdf_text() { local file="$1" # Extract text from PDF using pdftotext if command -v pdftotext &>/dev/null; then pdftotext "$file" - 2>/dev/null else echo "pdftotext not available; try PyPDF2 via Python sandbox" return 1 fi }
extract_raw_strings() { local file="$1" # Extract printable strings from any binary file if command -v strings &>/dev/null; then strings "$file" 2>/dev/null | head -500 else echo "strings not available" return 1 fi }
case "$1" in docx_raw) extract_docx_raw "$2" ;; xlsx_csv) extract_xlsx_to_csv "$2" ;; pdf_text) extract_pdf_text "$2" ;; raw_strings) extract_raw_strings "$2" ;; *) echo "Usage: $0 <docx_raw|xlsx_csv|pdf_text|raw_strings> <file>" exit 1 ;; esac
还没有评价。欢迎分享你的使用感受。
无需登录即可发表。欢迎分享使用感受、提问，或说明在哪些环境下无法运行。
还没有评论。来写第一条吧。
Prepare audio stems, the separate parts of a track, for mixing by matching their durations with silence, loops, or crossfades. Verify each generated and processed file before moving to the next production step.
中文简介正在准备中。当前显示英语简介。
Produce audio in separate stems, meaning individual instrument tracks, and verify each output as you go. Align track lengths with silence, loops, or crossfades, then apply effects, mix a master, and package the files.
中文简介正在准备中。当前显示英语简介。
Build serverless endpoints that relay external API requests for a dashboard while keeping API keys on the server. Includes Vercel examples for input checks, caching, cross-origin access, and consistent JSON responses.
中文简介正在准备中。当前显示英语简介。
Produce a master track and separate audio parts through a structured Python workflow. Apply filters and reverb, export WAV files with consistent settings, package them in a ZIP archive, and check that the deliverables meet the requested specifications.
中文简介正在准备中。当前显示英语简介。
Keep document creation moving when web searches and page retrieval repeatedly fail. The skill preserves the original task, uses the model’s existing knowledge, and clearly identifies information that could not be checked against external sources.
中文简介正在准备中。当前显示英语简介。
When Python execution repeatedly fails in execute_code_sandbox, use a file-based alternative: save the script, run it through the shell, and capture results or errors. The guidance covers paths, imports, timeouts, structured output, and optional cleanup.
中文简介正在准备中。当前显示英语简介。