You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Copy file name to clipboardExpand all lines: _posts/2026-05-13-executable-markdown-specs-agentic-coding.md
+12-12Lines changed: 12 additions & 12 deletions
Display the source diff
Display the rich diff
Original file line number
Diff line number
Diff line change
@@ -8,15 +8,15 @@ title: BDD for Spec-driven Agentic Development
8
8
context: Lightning Talk
9
9
---
10
10
11
-
11
+
12
12
13
13
Agentic software is changing how we build.
14
14
15
15
As agentic coding becomes more common, **how good your prompts specify the behaviour you want is a key to good agent dev outcomes**. The goal is longer autonomous execution loops: give an agent intent, let it plan, let it build, let it run checks, and keep going.
16
16
17
17
The workflow many of us are moving toward is:
18
18
19
-
> **Spec as prompt → Agent Plan → Agent Build → Run Automated Tests**
19
+
> **Prompt with a Spec → Agent Plans → Agent Builds → Agent Runs Tests**
20
20
21
21
Weak specs lead to unreliable autonomy. Strong, executable specs make autonomy safer.
22
22
@@ -30,10 +30,10 @@ But the key constraint remains intent.
30
30
31
31
The emerging workflow is simple:
32
32
33
-
1.**Spec as prompt**
34
-
2.**Agent Plan**
35
-
3.**Agent Build**
36
-
4.**Run Automated Tests**
33
+
1.**Prompt with a Spec**
34
+
2.**Agent Plans**
35
+
3.**Agent Builds**
36
+
4.**Agent Runs Tests**
37
37
38
38
BDD plays at both ends of this loop: examples help specify the behaviour in step 1, and automated tests provide the feedback in step 4. That feedback loop is what makes it durable and valuable over a long period.
39
39
@@ -111,18 +111,18 @@ That loud failure is the point. It tells humans and agents that either the imple
111
111
112
112
## 5. BDD + AI = spec-driven development loop
113
113
114
-
114
+
115
115
116
116
Put together, the practical workflow becomes:
117
117
118
-
1.**Spec as prompt**
119
-
2.**Agent Plan**
120
-
3.**Agent Build**
121
-
4.**Run Automated Tests**
118
+
1.**Prompt with a Spec**
119
+
2.**Agent Plans**
120
+
3.**Agent Builds**
121
+
4.**Agent Runs Tests**
122
122
123
123
This is where BDD and AI fit together naturally.
124
124
125
-
BDD is doing work at step 1 and step 4: it turns behaviour into examples up front, then turns those examples into tests that keep the loop honest. The spec gives the agent intent. The agent plan gives it structure. The agent build changes the system. Running automated tests closes the feedback loop. That is what makes the specification durable over time.
125
+
BDD is doing work at step 1 and step 4: it turns behaviour into examples up front, then turns those examples into tests that keep the loop honest. The spec gives the agent intent. The agent plans the work, builds the change, and runs tests to close the feedback loop. That is what makes the specification durable over time.
0 commit comments