The Agent Skills Directory

[SAFE]: The skill's primary function is instructional, providing documentation and examples for task decomposition. It does not automate any dangerous system actions or provide executable scripts.
[SAFE]: Code snippets provided in the documentation rely on standard, well-maintained libraries (dspy, pydantic). No untrusted external dependencies, remote script downloads, or unsafe execution patterns (like curl | bash) are present.
[PROMPT_INJECTION]: The skill demonstrates processing of untrusted external data such as medical reports, invoices, and resumes. While this constitutes an indirect prompt injection surface, there are no exploitable capabilities (such as command execution or file writes) that could be triggered by such data.
Ingestion points: 'report' in MedicalReportExtractor (examples.md), 'invoice_text' in InvoiceExtractor (examples.md), and 'resume_text' in ResumeParser (examples.md).
Boundary markers: Absent; inputs are passed directly to signatures.
Capability inventory: None; the skill performs structured text extraction only.
Sanitization: Absent.
[SAFE]: Metadata and instructions are consistent with the stated purpose of the skill. No obfuscation, persistence mechanisms, or privilege escalation patterns were detected.

ai-decomposing-tasks