The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests large volumes of untrusted content from external URLs and local files to extract brand voice traits, creating a vulnerability surface where embedded malicious instructions could influence agent behavior. 1. Ingestion points: Content is retrieved via WebFetch for URLs, Glob and Read for local files, and WebSearch for public content as described in SKILL.md. 2. Boundary markers: The instructions do not specify the use of delimiters or 'ignore embedded instructions' prompts to isolate external data from the system prompt. 3. Capability inventory: The skill has access to tools including Bash, Write, Edit, WebFetch, and WebSearch, which could be exploited if an indirect injection occurs. 4. Sanitization: There is no mention of sanitization, filtering, or validation of the content gathered from external sources before it is processed by the model.

brand-voice-analyzer