The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied financial data (assets, debts, and property details) to generate balancing worksheets and draft agreement clauses. This creates an ingestion surface for potential instructions embedded in user-supplied text.
Ingestion points: SKILL.md (Step 3), references/izun-mashabim-worksheet.md, and direct user interaction.
Boundary markers: The skill includes robust legal disclaimers at the top of the main instruction files (SKILL.md, SKILL_HE.md) and explicitly states that the agent is a navigator, not a lawyer.
Capability inventory: The skill is limited to text generation, reading its own bundled reference files, and executing the local scripts/izun_mashabim.py script. No network or high-privilege operations are enabled.
Sanitization: Not explicitly implemented in the provided Python calculation script.
[DYNAMIC_EXECUTION]: The skill instructions direct the agent to execute a bundled Python script (scripts/izun_mashabim.py) to perform arithmetic for property balancing.
Evidence: SKILL.md (Step 3) and scripts/izun_mashabim.py.
The script is local, static, and uses only standard Python libraries (json, argparse, sys). It does not perform network operations or access sensitive system files.
[METADATA_POISONING]: The skill cross-references another skill, israeli-pension-advisor, for actuarial calculations.
Evidence: SKILL.md (Step 3) and metadata.json.
This is a legitimate integration between skills authored by the same vendor (skills-il).

israeli-divorce-navigator