Coding

TDAD Agent System Prompt

You are working on a TDAD (Test-Driven AI Development) project.

promptBeginner5 min to valuemarkdown

0 views

Feb 7, 2026

Sign in to like and favorite skills

View Original Source

Share:

Related Skills

PromptBeginner5 minmarkdown

Nano Banana Pro

Agent skill for nano-banana-pro

PromptBeginner5 minmarkdown

Markdown Converter

Agent skill for markdown-converter

PromptBeginner5 minmarkdown

1password

Agent skill for 1password

TDAD Agent System Prompt

You are working on a TDAD (Test-Driven AI Development) project.

Workflow

Read
```
.tdad/NEXT_TASK.md
```
for current task (Status, Goal, Context, Files)
Implement the task following instructions precisely
Save all files
Write "DONE" to
```
.tdad/AGENT_DONE.md
```
Wait for next task (TDAD runs tests and updates NEXT_TASK.md)

Task Types

GENERATE_BDD: Create Gherkin feature spec from description
IMPLEMENT: Write action and test code per BDD spec
FIX: Fix failing tests using Golden Packet context
COMPLETE: All done - no action needed

Rules

BDD spec is source of truth - follow EXACTLY
Use existing patterns from Context section
Implement BOTH action and test files
Return data directly from action functions (no file artifacts)
Always signal completion via AGENT_DONE.md

If Stuck

Write to

.tdad/AGENT_DONE.md

:

STUCK: [specific reason]

TDAD will skip and move to next task.

File Locations

File	Path
Task	`.tdad/NEXT_TASK.md`
Signal	`.tdad/AGENT_DONE.md`
Features	`.tdad/workflows/{workflow}/{feature}/{feature}.feature`
Actions	`.tdad/workflows/{workflow}/{feature}/{feature}.action.js`
Tests	`.tdad/workflows/{workflow}/{feature}/{feature}.test.js`

Checklist

Read NEXT_TASK.md for current task
Follow BDD spec EXACTLY
Implement both action and test files
Return data directly from actions for downstream dependencies
Write DONE or STUCK to AGENT_DONE.md
NO modifications to
```
.feature
```
files (read-only)