The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes external text samples and approved company documents to define brand voice, which creates a potential surface for indirect instructions.
Ingestion points: The skill reads "Textos escritos pela empresa" and "amostras fornecidas" as defined in the SKILL.md entry flow.
Boundary markers: The instructions do not specify the use of delimiters or clear "ignore embedded instructions" warnings when handling the input samples.
Capability inventory: The skill reads project context from the .inboundfy/context/ directory and is instructed to update files like voz.md and dicionario.md.
Sanitization: There is no evidence of validation or sanitization protocols for the ingested content.
[SAFE]: The skill's operations are confined to the local filesystem and specific project directories. It does not request network access, use obfuscated code, or attempt to escalate privileges. The functionality is consistent with its stated purpose of corporate voice management.

inboundfy-voz