GuardrailsJailbreak DetectionSafetyAI Agent Skill
prompt-guard
Detects and blocks jailbreak and prompt-injection attempts before they reach the main model.
Install Options
Install into Claude Code, Cursor, Codex, or Antigravity with:
$ npx skills add orchestra-research/ai-research-skills@prompt-guard
Real Work Situation Solved
Audit and guardrail against critical hallucinations before public release
“The AI occasionally hallucinates fake promo policies or incorrect numbers - how do you install safety guardrails?”
5-Step Action Framework
- 1List mission-critical facts that must never be wrong (pricing, legal terms, refund policies, PII).
- 2Deploy input guardrails: detect and reject jailbreak attempts and adversarial prompt injection.
- 3Deploy output verification: use regex and rule validation to cross-check numbers, links, and promo codes against the database.
- 4Define a deterministic fallback: when model confidence drops or checks fail, redirect to safe canned responses or human support.
- 5Log 100% of guardrail interventions into a review dashboard to feed ongoing eval benchmark updates.
Before
"Add a system prompt note: 'always answer accurately, never make things up'."
After
"Added a deterministic post-check: all generated prices must match the database exactly; on any mismatch, intercept and output official pricing link."
Related Course Lesson
AI Agents for PM • Lesson 9
SKILL.md Source Instructions
SKILL.md • Read-only preview# Prompt Guard Skill Screen incoming prompts against known jailbreak and injection patterns, and route flagged input to a safe fallback instead of the primary model.