The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill acts as an orchestrator that ingests potentially untrusted data from user-provided business ideas and dossier sections, creating a surface for indirect prompt injection.
Ingestion points: Untrusted data enters the agent context through the "source business idea" or "note path" specified in references/section-generator.md and via individual dossier sections referenced in SKILL.md.
Boundary markers: The skill explicitly implements "Section Isolation Rules" and a "Non-Negotiable Rule" in references/dossier.md and references/orchestrator.md. These rules attempt to prevent cross-section contamination by requiring each section to be generated and reviewed in an isolated, closed loop.
Capability inventory: The skill possesses the capability to write the final assembled dossier to a filesystem path provided by the user, as described in references/dossier.md.
Sanitization: The provided instructions and reference files do not show evidence of explicit content sanitization or instruction filtering for the ingested data.

write-dossier