The Agent Skills Directory

[Hardened YAML Parsing]: The validation tool scripts/validate_feasibility.py uses a custom loader that inherits from yaml.SafeLoader. This implementation prevents the instantiation of arbitrary Python objects during parsing, mitigating risks associated with unsafe deserialization. The loader also explicitly rejects YAML anchors and aliases to prevent resource exhaustion attacks.
[File System Access Controls]: The script implements validation logic to prevent path traversal attacks by rejecting '..' segments and ensuring files are within permitted roots. Additionally, it enforces a 5MB size limit on input files to protect against denial-of-service attempts.
[Input Ingestion Surface]: The skill processes user-supplied Markdown studies at runtime. While this presents an indirect prompt injection surface, the risk is addressed through the use of specific boundary markers ( tags) to delimit data and narrative sections, alongside schema validation using jsonschema to ensure data integrity.

ds-feasibility