Skip to content

docs: add prompt hidden content sanitization gates#1391

Open
catcherintheroad-hub wants to merge 1 commit into
UnitOneAI:mainfrom
catcherintheroad-hub:improve/prompt-hidden-content-sanitization
Open

docs: add prompt hidden content sanitization gates#1391
catcherintheroad-hub wants to merge 1 commit into
UnitOneAI:mainfrom
catcherintheroad-hub:improve/prompt-hidden-content-sanitization

Conversation

@catcherintheroad-hub
Copy link
Copy Markdown

Created from review issue: #1390

Summary

  • Add hidden content extraction and sanitization evidence gates to the prompt-injection skill.
  • Require loader behavior evidence for HTML, markdown, PDF/office documents, email/message content, and tool/API response metadata.
  • Add an External Content Extraction Review table and edge-case fixtures for hidden HTML, markdown link exfiltration, PDF/OCR metadata, and field-level provenance.

Validation

  • git diff --check
  • Markdown fence/non-ASCII sanity check
  • Reference URL checks for OWASP LLM01, OWASP LLM Top 10, and Greshake et al. indirect prompt injection paper

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant