The Agent Skills Directory

[SAFE]: The skill explicitly defines a 'Content Trust Boundary' and provides clear instructions across its operations (extract.md, audit.md, critique.md) to treat all external data as inert facts, mitigating indirect prompt injection risks from processed URLs and documents.
[SAFE]: Local Python scripts (slop_scan.py and validate_copy.py) perform deterministic scanning using safe methods such as yaml.safe_load() and regex-based leakage detection without performing network or privileged system operations.
[SAFE]: The use of WebFetch for URL content extraction is directly tied to the primary purpose of the skill and is governed by strict prompts to ignore embedded directives in source material.

copywriting