The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill possesses an indirect prompt injection surface as it ingests untrusted external inputs, such as competitor URLs, third-party codebases, and project documentation.
Ingestion points: Processes external files including .jez/artifacts/research-brief-{topic}.md, CLAUDE.md, user briefs, and remote competitor URLs.
Boundary markers: No specific boundary markers or instructions are defined to separate untrusted external text from the execution instructions.
Capability inventory: Employs powerful capabilities via allowed tools including Bash (e.g., npx wrangler deploy), Write, and Edit to perform long-running autonomous development pipelines.
Sanitization: Lacks explicit validation or sanitization mechanisms for the data parsed from files or external webpages before feeding it into the agent context.

roadmap