PromptGuard can grow into a focused terminal UI for reviewing prompt contracts.
- show findings without dumping long logs
- make missing decisions obvious
- provide rewrite drafts
- let users accept risks intentionally
- support repo-level audits and pasted prompt audits
┌─ PromptGuard ───────────────────────────────────────────────────────────────┐
│ Source: prompts.py Findings: 4 Blocking: 2 │
├───────────────────────────────┬─────────────────────────────────────────────┤
│ Findings │ Detail │
│ │ │
│ HIGH PG012 responsibility │ Evidence │
│ HIGH PG015 technical risk │ "Fix this bug and write code..." │
│ MED PG005 output contract │ │
│ LOW PG006 maintainability │ Missing Contract │
│ │ Role, owned surface, verification, │
│ │ changed-files report, residual risk. │
│ │ │
│ │ Questions To Ask │
│ │ 1. Which module/files are owned? │
│ │ 2. What behavior must be preserved? │
│ │ 3. How should success be verified? │
├───────────────────────────────┴─────────────────────────────────────────────┤
│ [F] Fix draft [A] Accept risk [S] Save report [Q] Quit │
└─────────────────────────────────────────────────────────────────────────────┘
j/k: move between findingsenter: expand finding detailf: copy or write fix drafta: accept risk with reasons: save report to.promptguard/reports.jsonl/: filter by rule id or severity
- dark terminal background
- dense two-pane review layout
- severity badges using red/yellow/blue
- no decorative dashboard cards
- optimized for repeated engineering use