pm-research-about

Pass

Audited by Gen Agent Trust Hub on Feb 26, 2026

Risk Level: SAFE
Full Analysis
  • [PROMPT_INJECTION]: The skill identifies web content as untrusted and includes explicit instructions to ignore any commands found within fetched data, focusing solely on fact extraction to prevent prompt injection attacks.
  • [COMMAND_EXECUTION]: Shell command access is narrowly restricted to git add and git commit for maintaining the research index, preventing arbitrary command execution.
  • [DATA_EXFILTRATION]: While the skill uses network tools (WebFetch, WebSearch), it includes mandatory rules against revealing secrets or following instructions from external URLs, effectively mitigating exfiltration risks during the research process.
  • [EXTERNAL_DOWNLOADS]: The skill references local context and template files. It fetches data for research purposes but does not download or execute remote scripts or packages.
Audit Metadata
Risk Level
SAFE
Analyzed
Feb 26, 2026, 08:48 PM