plan-then-execute
Frozen plans with human approval gates.
When & Why to Use This Skill
The 'plan-then-execute' skill provides a robust security framework for AI agents by enforcing a two-phase workflow: isolated planning followed by human-approved execution. It effectively mitigates prompt injection and redirection attacks by freezing the action sequence before untrusted data is processed, ensuring the agent remains within safe, predefined boundaries and follows a human-verified path.
Use Cases
- Secure Code Deployment: Ensuring that deployment scripts and terminal commands are reviewed and frozen before execution, preventing malicious logs or tool outputs from hijacking the release process.
- Untrusted Data Processing: Safely extracting information from external documents or websites by defining the tool sequence before the agent interacts with potentially malicious or injected content.
- Sensitive System Operations: Managing infrastructure or database changes where every step must be explicitly approved by a human operator to prevent unintended or unauthorized modifications caused by dynamic tool feedback.
| name | plan-then-execute |
|---|---|
| description | "Frozen plans with human approval gates." |
| license | MIT |
| tier | 1 |
| related | [planning, action-queue, representation-ethics] |
| tags | [moollm, execution, security, approval, frozen] |
Plan Then Execute
Frozen plans with human approval gates.
Two-phase execution: plan in isolation, execute the frozen sequence.
[!CAUTION] Security first. Tool outputs cannot alter the plan. Human approval required.
Why This Exists
If tool outputs can alter the choice of later actions, injected instructions may redirect the agent toward malicious steps. This skill enforces:
- Plan phase — Generate tool sequence before seeing untrusted data
- Approval gate — Human reviews and approves
- Execution phase — Run exactly that sequence
Contents
| File | Purpose |
|---|---|
| SKILL.md | Full protocol documentation |
| PLAN.yml.tmpl | Plan template |
| EXECUTION_LOG.md.tmpl | Execution log template |
Quick Example
# PLAN.yml
plan:
name: "Deploy to staging"
status: approved # Frozen after approval
steps:
- id: 1
name: "Run tests"
tool_call:
tool: "terminal.run"
args: { command: "npm test" }
status: pending
The Intertwingularity
Plan-then-execute is planning with security guarantees.
graph LR
PTE[📋 plan-then-execute] -->|frozen variant of| PL[🗂️ planning]
PTE -->|logs to| SL[📜 session-log]
PTE -->|can use| TC[🎴 card]
SS[👯 sister-script] -->|produces| PTE
Dovetails With
Sister Skills
| Skill | Relationship |
|---|---|
| planning/ | Flexible, evolving alternative |
| session-log/ | Execution gets logged |
| sister-script/ | Scripts become plans |
Protocol Symbols
| Symbol | Link |
|---|---|
PLAN-EXECUTE |
PROTOCOLS.yml |
APPEND-ONLY |
PROTOCOLS.yml — Execution log |
WHY-REQUIRED |
PROTOCOLS.yml — Every step explains intent |
Navigation
| Direction | Destination |
|---|---|
| ⬆️ Up | skills/ |
| ⬆️⬆️ Root | Project Root |
| 🗂️ Sister | planning/ |