The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from user descriptions, changelogs, and documentations to update persistent context files in the .inboundfy/context/ directory. These files are used as a source of truth for downstream processes, which could lead to indirect prompt injection if malicious instructions are included in the source data.
Ingestion points: User input and raw materials are processed in SKILL.md and references/produtos.md to update files like produtos.md, servicos.md, and ofertas.md.
Boundary markers: The skill uses Markdown templates, headers, and YAML frontmatter in REFERENCIA.md to structure data, which provides some structural separation but lacks explicit instructions to disregard or escape embedded commands.
Capability inventory: The skill is capable of reading and writing Markdown and YAML files within the project's .inboundfy/ directory structure.
Sanitization: Data is processed through an extraction and classification step performed by the AI, which acts as a logical filter; however, there is no technical sanitization, escaping, or strict schema validation for the content before it is written to the file system.

inboundfy-contexto-oferta