tweet-draft-reviewer

Pass

Audited by Gen Agent Trust Hub on Mar 8, 2026

Risk Level: SAFE
Full Analysis
  • [COMMAND_EXECUTION]: The skill uses a bash script with find and grep to search for unreviewed markdown files in the local content/tweet-drafts/ directory. This is a routine operation for the skill's batch-processing feature.\n- [PROMPT_INJECTION]: The skill exhibits an attack surface for indirect prompt injection because it processes content from external files.\n
  • Ingestion points: Markdown files in the content/tweet-drafts/ folder and direct user-pasted text.\n
  • Boundary markers: None; the skill does not explicitly instruct the agent to ignore instructions contained within the draft text.\n
  • Capability inventory: Restricted to local file discovery and reading via the bash tool. No network access, file writing, or administrative commands are present.\n
  • Sanitization: None; draft content is processed directly to evaluate style rules.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 8, 2026, 11:30 PM