The Agent Skills Directory

[PROMPT_INJECTION]: The skill instructions do not contain any patterns attempting to bypass safety filters, override system prompts, or extract instructions. It focuses on drafting documents based on user-provided data.
[DATA_EXPOSURE_AND_EXFILTRATION]: No sensitive data, hardcoded credentials, or exfiltration patterns were found. Network references are limited to public, well-known legal and information resources in Israel (e.g., justice.gov.il, kolzchut.org.il).
[OBFUSCATION]: No Base64, zero-width characters, homoglyphs, or encoded payloads were detected. The Hebrew content is standard text.
[REMOTE_CODE_EXECUTION]: No remote scripts are downloaded or executed. The skill operates using local Markdown templates and a small, static Python script.
[PRIVILEGE_ESCALATION]: The skill does not use administrative commands, sudo, or system-level permission modifications.
[PERSISTENCE_MECHANISMS]: The skill does not attempt to modify shell profiles, scheduled tasks, or startup folders.
[METADATA_POISONING]: Metadata fields (author, description, tags) are consistent with the skill's stated purpose and do not contain hidden instructions.
[INDIRECT_PROMPT_INJECTION]: While the skill ingests user input (such as names and assets) to interpolate into templates, it lacks dangerous capabilities (network sends, file writes, or command execution) that could be exploited via malicious input. Ingestion points are confined to simple identity and asset strings for text generation.
[TIME_DELAYED_ATTACKS]: No time-based or environment-based logic gates were found that would trigger suspicious behavior at a later date.
[DYNAMIC_EXECUTION]: The provided script scripts/inheritance_helper.py uses standard string parsing for basic logic and does not utilize eval(), exec(), or dynamic path loading. It performs deterministic checks on witness eligibility and order types.

israeli-wills-inheritance