The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses a suite of local command-line utilities (e.g., inboundfy doctor, inboundfy acervo add, inboundfy-setup) to manage project configuration and file registration. These tools are consistent with the vendor's specialized framework for content management.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data such as website links, narrated text, and documents, creating a potential surface for indirect prompt injection. However, the skill architecture includes high-quality guardrails:
Ingestion points: External content enters through the 'Entrada esperada' section defined in SKILL.md.
Boundary markers: Original inputs are preserved in an immutable bruto.md file, and the workflow requires seven distinct 'anti-slop' auditing milestones (A0-A6) to maintain integrity across versions.
Capability inventory: Capabilities include managed local file writes and web research conducted through vendor-specific tools (inboundfy-pesquisa-acervo).
Sanitization: Content is filtered through a project-specific glossary and prohibited terms list, with mandatory human checkpoints for any factual claims or editorial decisions.

inboundfy-acervo