The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process untrusted external data (raw transcriptions, notes, and research material) as its primary function, which presents an attack surface for indirect prompt injection.
Ingestion points: references/etapas/00-triagem.md specifically identifies the ingestion of raw, unedited material such as meeting transcriptions and informal briefings.
Boundary markers: The framework utilizes rigid Markdown templates and YAML frontmatter to delimit data, although it does not explicitly use "ignore embedded instructions" delimiters for the raw content during processing.
Capability inventory: The skill has the capability to write files across the project structure and orchestrate other specialized skills (inboundfy-*). references/etapas/02-pesquisa.md also suggests performing external research via URLs.
Sanitization: The skill includes a dedicated stage (references/etapas/01-saneamento.md) for cleaning and organizing material, and utilizes an "anti-slop" auditor and validation skills at multiple milestones to ensure output integrity and adherence to context guidelines.
[SAFE]: The skill follows professional development standards, utilizing isolated stages for triage, sanitization, research, planning, briefing, production routing, and final auditing. It enforces strict adherence to local context files (.inboundfy/context/) and does not perform any sensitive system operations, privilege escalation, or persistence attempts.

inboundfy-planejamento