The Agent Skills Directory

[PROMPT_INJECTION]: The skill acts as a manual for circumventing content safety filters.
[PROMPT_INJECTION]: It promotes 'The Chinese Prompt Trick' as a method to evade safety thresholds by switching languages.
[PROMPT_INJECTION]: It provides a 'High-Risk Word Categories' table that maps sensitive terms like 'kill' or 'sensual' to 'Safer Alternatives' intended to mask policy-violating intent from the model.
[PROMPT_INJECTION]: It includes 'Production Context Tokens' (referred to as a 'Filter Shield') which are keywords designed to manipulate the model's safety evaluation and increase the pass rate for restricted content.
[NO_CODE]: No executable scripts or code files were detected within the skill package.

seedance-filter