Skills
skills/mpuig/agent-slides/slides-extract/Gen Agent Trust Hub

slides-extract

Pass

Audited by Gen Agent Trust Hub on Mar 17, 2026

Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill executes various commands using uvx, such as slides extract, slides docs, and slides preflight, to process PowerPoint templates and verify the generated project structure.\n- [EXTERNAL_DOWNLOADS]: The skill uses uvx --from agent-slides, which triggers the download of the agent-slides package from a Python package registry at runtime.\n- [REMOTE_CODE_EXECUTION]: By using uvx --from agent-slides, the skill dynamically fetches and runs the agent-slides utility, which constitutes execution of external code.\n- [PROMPT_INJECTION]: The skill processes user-provided .pptx files and uses the resulting extracted data to build a design profile, creating a vulnerability surface for indirect prompt injection.\n
  • Ingestion points: User-provided .pptx templates and the generated resolved_manifest.json file (SKILL.md).\n
  • Boundary markers: There are no explicit instructions or delimiters used to separate the processed content from the agent's instructions or to ignore potentially malicious embedded text.\n
  • Capability inventory: The skill has access to shell execution via uvx and file system management via mkdir (SKILL.md).\n
  • Sanitization: No validation, escaping, or sanitization steps are defined for the PowerPoint content or the extracted manifest data before it is used in subsequent commands.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 17, 2026, 06:32 AM