court-form-filling-manus-familytools-fullinstr-v2-skill

name	court-form-filling-manus-familytools-fullinstr-v2-skill
description	Family-scoped SkillLearnBench skill for court-form-filling using latest full-instruction v2 family injection.
allowed-tools	["date_normalize","list_dir","pdf_extract_text","pdf_field_value_dump","pdf_fill_form_fields","pdf_list_form_fields","pdf_render_pages","read_file","run_python","run_shell","write_file"]
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"}

Skill: Court Form Filling

When to use this skill

Apply this skill for court-form-filling tasks in SkillLearnBench. The goal is to fill a legal PDF form (e.g., California Small Claims) based on a case description, mapping facts to fields and saving to a specific path.

Visible input and artifact inventory

• Inputs: Task instruction (case facts, constraints like date formats), blank PDF form (e.g., /root/sc100-blank.pdf), and config files (instruction.md, task.toml).
• Artifacts: A filled PDF form saved to the exact path requested (e.g., /root/sc100-filled.pdf).

Execution procedure for the current task

1. Inventory: Read the instruction and locate the blank PDF and required output path.
2. Discovery: Run pdf_list_form_fields on the blank PDF to get exact field names.
3. Mapping: Extract facts (names, addresses, amounts, dates) from the instruction. Map them to discovered fields. Leave unmentioned or court-only fields empty.
4. Transformation: Format dates as requested (e.g., xxxx-xx-xx) using date_normalize.
5. Filling: Use pdf_fill_form_fields to populate the PDF and save it to the target path.
6. Verification: Run pdf_field_value_dump on the output PDF to verify filled values.

Family tool routing and useful placeholder snippets

Prefer family-specific tools over ad hoc shell scripts:

• pdf_list_form_fields: Discover field names.
• pdf_fill_form_fields: Generate the artifact.
• pdf_field_value_dump: Verify contents.
• date_normalize: Format dates.

Snippet:

# Map extracted facts to discovered fields
fields = {"Plaintiff_Name": "Extracted Name", "Claim_Amount": "1500"}
# Use pdf_fill_form_fields(input_pdf, output_pdf, fields)

Validation checks before final submission

• Path: Ensure the filled PDF exists at the exact requested output path.
• Content: Check pdf_field_value_dump output to confirm facts are correctly mapped.
• Constraints: Verify date formats and ensure unmentioned fields remain blank.

Common failure modes and repair actions

• Bad Field Names: If filling fails, re-run pdf_list_form_fields to correct the mapping keys.
• Format Mismatch: If dates are wrong, review the instruction and reformat before filling.
• Overfilling: If court fields are filled, clear their mapping and regenerate the PDF.

Finalizer capsule

If near the step limit without a perfect artifact:

1. Extract the most critical facts (Plaintiff, Defendant, Amount).
2. Map them to the most obvious fields found via pdf_list_form_fields.
3. Run pdf_fill_form_fields to create the file at the required path. A partially correct file at the right path is better than none.