| name | dou |
| description | Date-stamped iterative goal execution across 6 passes (ITERATION 0 through 5). Use when the user says ">doU" or "run doU", asks for iterative refinement with tracking, wants to execute a prompt through multiple progressive passes, or needs a goal file with iteration history. Triggers on any request involving iterative execution with date-stamped goal files. |
| license | MIT |
| compatibility | No dependencies. Works with any agent that can execute shell commands and write files. |
| metadata | {"author":"nbiish","version":"1.0.0","trigger":">doU","iterations":"6","spec":"agentskills.io"} |
doU — Date-Stamped Iterative Goal Execution
Execute a prompt through 6 progressive refinement passes, stamping each into a date-named goal file.
The Expansion
When you encounter >doU, the user's prompt goes into the placeholder:
>doU """{PROMPT}"""
If the user provides the prompt inline without triple quotes, treat their entire message after >doU as {PROMPT}.
Execution Protocol
Step 1: Initialize the goal file
DATE=$(date +"%b %d, %Y")
TIMESTAMP=$(date)
GOAL_FILE="goal-${DATE}.md"
Create or append to goal-{date}.md in the current working directory:
# ITERATION 0
{TIMESTAMP}
Step 2: Execute iterations 0 through 5
For each iteration N (starting at 0):
-
Execute the prompt fully, producing complete output
-
Append the output under the current # ITERATION N header in the goal file
-
Reflect on the output — what was done, what gaps remain, what the next pass should improve
-
If N < 5: Stamp the next iteration header and re-execute with accumulated learning:
# ITERATION {N+1}
{new timestamp from `date`}
Re-execute the same prompt, but incorporate all improvements and corrections identified through iterations 0–N.
-
If N = 5: Produce the final production-grade output, then append:
<!-- doU complete: 6 iterations -->
Step 3: Report completion
After ITERATION 5:
doU complete. 6 iterations executed.
Goal file: goal-{date}.md
Iteration Focus
Each pass has a specific purpose. The agent should lean into the appropriate focus for each iteration:
| Pass | Focus | What to emphasize |
|---|
| 0 | Initial execution | Produce working, complete output. Raw but functional. |
| 1 | Correction | Fix errors, fill gaps, address anything broken or missing from pass 0. |
| 2 | Refinement | Improve structure, naming, completeness, readability. Tighten the work. |
| 3 | Optimization | Performance, edge cases, error handling, robustness. |
| 4 | Hardening | Security, validation, constraints, compliance, production readiness. |
| 5 | Final review | Production-grade. Polished, documented, no loose ends. |
Output Contract
- File:
goal-{date}.md in the current working directory
- Format: 6 iteration sections, each with a
# ITERATION N header, timestamp from date, and the output
- Quality gradient: Each iteration must demonstrate measurable improvement over the previous one
- Append-only: Never modify previous iterations. Each pass appends only.
Constraints
- All 6 iterations must execute. Do not skip any.
- Each pass must add value. Do not repeat previous output verbatim.
- Each pass must be timestamped with the output of
date.
- If the prompt is empty, ask the user for a prompt before starting.
- Multiple doU calls on the same date append to the same goal file with distinct iteration chains.
Examples
Basic usage:
>doU """Design a REST API for task management"""
Inline (no triple quotes):
>doU analyze this codebase for security vulnerabilities
Orchestrator delegation:
>doU """{subagent_task}"""
Multiple Invocations
Running >doU twice in one session creates two iteration chains in the same date file:
# ITERATION 0
Mon May 26 14:32:01 EDT 2026
{first task output}
# ITERATION 1
...
<!-- doU complete: 6 iterations -->
# ITERATION 0
Mon May 26 15:01:22 EDT 2026
{second task output}
...
Each chain runs 0→5 independently.