| name | financial-analysis-manus-familytools-fullinstr-v2-skill |
| description | Family-scoped SkillLearnBench skill for financial-analysis using latest full-instruction v2 family injection. |
| allowed-tools | ["extract_archive","list_dir","read_file","run_python","run_shell","search_text","sec13f_fund_details","sec13f_holdings_compare","sec13f_holdings_query","sec13f_index_search","sec13f_top_holders","write_file","write_json"] |
| metadata | {"benchmark":"SkillLearnBench","generator":"Manus","baseline":"manus_familytools_fullinstr_v2","generation_constraints":"slb_no_instance_memory_familytools_v2","allowed_gold_input":"familytools_full_instruction_v2","skill_injection_scope":"family_scoped","date":"2026-05-23"} |
Financial Analysis Skill
When to use this skill
Use this skill for tasks analyzing hedge fund activities, such as comparing financial data across quarters (e.g., Q2 and Q3). It applies when determining Assets Under Management (AUM), stock counts held by funds, identifying stocks with increased investments, or ranking fund managers by investments in specific stocks using SEC 13F filings.
Visible input and artifact inventory
The environment typically includes:
- Zipped SEC 13F filings datasets (e.g.,
13f-2025-q2.zip, 13f-2025-q3.zip).
- An
instruction.md detailing questions.
- A
task.toml with metadata.
- A
Dockerfile for setup.
The expected output is a JSON file named answers.json in /root, formatted per the instructions' schema.
Execution procedure for the current task
- Initial Setup: Read
instruction.md for questions and output format. Inventory visible files to locate datasets.
- Data Extraction: Use
extract_archive to unzip datasets into directories (e.g., /root/2025-q2).
- Fund Identification: For fund-specific data (AUM, stock count), use
sec13f_index_search to fuzzy search the COVERPAGE for the fund's name to get its accession_number.
- Data Retrieval:
- Use
sec13f_fund_details with the accession_number for AUM and stock count.
- For comparing holdings, get the
accession_number for both quarters. Use sec13f_holdings_compare to analyze changes and identify top stocks with increased investments.
- For top fund managers investing in a stock, find the stock's CUSIP, then use
sec13f_top_holders to rank funds by share value.
- Artifact Creation: Compile answers into a JSON object matching the schema. Use
write_json to save it to /root/answers.json.
Family tool routing and useful placeholder snippets
extract_archive: Unzip datasets.sec13f_index_search: Find accession_number by searching COVERPAGE.
accession_number = sec13f_index_search(search_term="fund name", quarter="2025-q3")
sec13f_fund_details: Get AUM and stock count.
details = sec13f_fund_details(accession_number="0001234567-25-000001")
sec13f_holdings_compare: Compare holdings between quarters.
comparison = sec13f_holdings_compare(accession_number_q2="...", accession_number_q3="...")
sec13f_top_holders: Find top investors in a stock.
top_holders = sec13f_top_holders(cusip="123456789", quarter="2025-q3")
write_json: Save final answers.
Validation checks before final submission
- Verify
answers.json exists in /root.
- Ensure JSON structure exactly matches the schema (e.g.,
q1_answer is a number, q3_answer is a list).
- Check all questions are answered with correct data types.
Common failure modes and repair actions
- Missing
accession_number: Try variations of the fund's name or check for typos.
- Incorrect JSON Schema: Review data types and keys used during
write_json and correct them.
- Data Extraction Errors: Ensure
extract_archive was used correctly and paths match instructions.
Finalizer capsule
If nearing the maximum step limit and analysis is incomplete, create a minimally compliant answers.json in /root. It should contain correct keys per the schema, even with default values (e.g., 0 for numbers, [] for lists), to ensure correct artifact shape and prevent complete failure.
