The Agent Skills Directory

[SAFE]: The skill operates primarily as a knowledge-only assistant. No malicious command execution or unauthorized network activity was found. The instructions provided are pedagogical and professional in nature.
[SAFE]: The provided Python script governance.py implements legitimate logging and rate-limiting functionality. It is restricted to managing a local data directory within the skill folder for audit trails and does not exhibit malicious filesystem or network behaviors.
[SAFE]: External references point to official Brazilian government and banking portals (such as Caixa Econômica Federal and Banco do Brasil) for property auctions. These are well-known and legitimate services in the context of the skill's purpose.
[SAFE]: The skill processes untrusted external data (auction notices), which represents a standard indirect prompt injection surface for this use case. However, it lacks high-privilege capabilities (like shell access or network exfiltration) that would make this surface exploitable.
Ingestion points: User-provided auction notices (editais) as defined in SKILL.md.
Boundary markers: Absent; the protocol relies on the agent's internal reasoning without explicit delimiters for input data.
Capability inventory: Local file writing for logging in scripts/governance.py.
Sanitization: None detected in the instructional or script logic.

leiloeiro-edital