Preflight prompt for unblocking AI agents before long autonomous sessions

agent-preflight-prompt.md

Agent Preflight Prompt

Paste this after your task brief to surface blockers before long autonomous sessions.

The Prompt

## PREFLIGHT CHECK - Do Not Start Work Yet

Before implementing anything, conduct a thorough preflight analysis. Surface every potential blocker NOW so we can resolve them before you begin.

### Step 1: Understand the Task
Restate the task in your own words so I can confirm alignment.

### Step 2: Explore the Codebase
- What files/modules will this touch?
- What patterns and conventions exist?
- What dependencies or integrations are involved?

### Step 3: Enumerate Blockers

For each category, list what you need or flag "None identified":

**🔑 Credentials & Access**
- API keys, tokens, secrets needed
- Database access or connection strings
- Third-party service authentication
- Deployment permissions

**🤔 Decisions I Need to Make**
- Architectural choices with multiple valid approaches
- UX/behavior decisions
- Trade-offs where you have a preference
- Naming/structure choices not covered by existing patterns

**📐 Scope Clarifications**
- Ambiguities in requirements
- Edge cases not addressed
- "Should I also..." questions
- Where exactly should I stop?

**🚧 Constraints & Boundaries**
- Can I install new dependencies?
- Can I modify shared code/utilities?
- Can I create new files/directories?
- Can I run destructive operations?
- Environment restrictions?

**⚠️ Assumptions I'm Making**
List anything you're assuming that, if wrong, would change your approach.

**🔗 Dependencies & Sequencing**
- Does anything need to happen first?
- Parallel workstreams that might conflict?
- Output needed from another tool/agent/human?

### Step 4: Summarize

1. **Confidence level** (High/Medium/Low) for autonomous completion once unblocked
2. **Blockers**: X items need resolution
3. **Critical path**: The ONE thing that would stop you fastest if unresolved

Do NOT begin implementation until I respond to your preflight report.

Usage

1. Give the agent your task brief
2. Paste the prompt above
3. Resolve everything the agent surfaces
4. Confirm and let it run

Example

I need you to implement user authentication with OAuth.

[paste preflight prompt]

The agent will analyze, explore, and enumerate blockers before starting work.

---

💡 Pro tip: Most agent tools support custom slash commands. Save this as /preflight in Claude Code (.claude/commands/), Cursor, Windsurf, or your tool of choice—then just type /preflight after any task brief.