draft-writing

Pass

Audited by Gen Agent Trust Hub on Mar 1, 2026

Risk Level: SAFE
Full Analysis
  • [SAFE]: The skill is purely functional, providing a logical workflow for academic writing, including section-specific guidelines and validation rules.
  • [COMMAND_EXECUTION]: All tools defined (write_draft, patch_draft, count_words) are task-specific and operate on the local file system within the project directory. There are no attempts to execute arbitrary shell commands or escalate privileges.
  • [DATA_EXFILTRATION]: No network-enabled tools or operations are present. The skill manages citations and drafts locally, with no evidence of data being sent to external domains.
  • [PROMPT_INJECTION]: The use of 'protected blocks' (🔒) is a functional design to ensure the agent maintains consistency with the research novelty and selling points. These are not attempts to override the AI's core safety guidelines.
  • [REMOTE_CODE_EXECUTION]: The skill does not download or execute external scripts. It references a local 'DraftSnapshotManager' and 'run_writing_hooks' which are part of the internal infrastructure for file versioning and text validation.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 1, 2026, 08:10 AM