The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted narrative data from users (founders) and uses it to generate file content for a generated "child skill".
Ingestion points: The skill requests personal/professional stories and company history as described in SKILL.md (Quick start, Dados necessários) and references/ENTREVISTA_FUNDADORES.md.
Boundary markers: Absent. The skill instructions do not specify the use of delimiters to isolate user narratives from agent instructions during processing.
Capability inventory: The skill has file creation capabilities (SKILL.md Step 10) and network access for research (SKILL.md Step 7).
Sanitization: Present. The skill is explicitly instructed to perform an editorial sanitization process to remove sensitive personal data (health, religion, politics, etc.) before generating the child skill, as detailed in SKILL.md (Step 9) and references/ENTREVISTA_FUNDADORES.md (Privacidade section).
[COMMAND_EXECUTION]: The skill instructions involve the creation of file directories and the generation of ZIP archives to deliver output.
Evidence: SKILL.md Step 10 ("Gerar a skill-filha") and Step 14 ("Entregar o diretório ou ZIP").
[EXTERNAL_DOWNLOADS]: The skill performs network operations to research public brand information.
Evidence: SKILL.md Step 7 ("Pesquisar apenas com autorização") and "Política de internet" section. The skill requires explicit user authorization before accessing external websites or social media.

skill-soul-builder-by-rdd