inbox-processing-expert

Automate PDF inbox processing by building a suggestion-based architecture that applies TOCTOU prevention, command-injection guards, and atomic writes. The Skill provides patterns for secure metadata extraction, confidence scoring, and idempotent workflows via SHA256 registries to ensure safe, repeatable processing.

TOCTOU (time-of-check-time-of-use) vulnerabilities in file automation are prevented through atomic writes, idempotent registries, and structured validation gates. This Skill covers command-injection prevention and engine/interface separation to harden inbox automation against race conditions and unsafe file operations.

Set up an idempotent inbox workflow by implementing a suggestion engine that scans PDFs, extracts metadata with confidence scoring, and logs actions atomically. Use SHA256 registries to track processed files, ensuring repeated runs produce consistent results without duplicate processing or state corruption.

Yes. AI-driven metadata extraction applies LLM detection and structured tagging to inbox attachments. This Skill implements confidence scoring and error taxonomy so you can validate extraction quality, filter low-confidence suggestions, and maintain audit trails for compliance.

Security-hardened file automation requires TOCTOU prevention, command-injection blocking, atomic writes, and structured logging. This Skill covers engine/interface separation, suggestion-based validation, and error taxonomy to build resilient, auditable systems that reject unsafe operations.

Implement suggestion-based workflows by separating the inbox engine from the interface layer, generating candidate actions with confidence scores, and presenting them for interactive approval. This pattern enables secure, auditable automation where each action is validated before execution.