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: website/docs/practical-techniques/lesson-7-planning-execution.md
+21-4Lines changed: 21 additions & 4 deletions
Display the source diff
Display the rich diff
Original file line number
Diff line number
Diff line change
@@ -117,7 +117,14 @@ Review the "why" behind the plan, not just the "what." If the agent says "Implem
117
117
118
118
**Example:** Agent proposes caching user sessions in Redis with 24-hour TTL. Good plan—but did it check your existing session implementation? Did it consider GDPR compliance for session data? Did it account for cache invalidation when users change passwords?
119
119
120
-
If grounding was shallow, stop and add context before execution. "Before implementing, research our existing session management, check for compliance requirements in COMPLIANCE.md, and propose cache invalidation strategy for security-critical events. Use ArguSeek, validate the approach against current security best practices."
120
+
If grounding was shallow, stop and add context before execution:
121
+
122
+
```markdown
123
+
Before implementing, research our existing session management,
124
+
check for compliance requirements in COMPLIANCE.md, and propose cache
125
+
invalidation strategy for security-critical events.
126
+
Use ArguSeek, validate the approach against current security best practices.
127
+
```
121
128
122
129
### Glance Over Suggested Changes
123
130
@@ -133,7 +140,12 @@ This is high-level architectural fit, not line-by-line code review (validation c
133
140
134
141
**Example:** Agent plans to add email validation by creating a new validation library in `src/lib/validators/`. But you already have Zod schemas in `src/validation/`. This is a grounding failure—the agent generated a plausible solution from training patterns instead of discovering your existing validation approach.
135
142
136
-
Stop and correct: "We use Zod for validation—check `src/validation/userSchema.ts` and follow that pattern instead of creating a new library."
143
+
Stop and correct:
144
+
145
+
```markdown
146
+
We use Zod for validation—check `src/validation/userSchema.ts` and follow that
147
+
pattern instead of creating a new library.
148
+
```
137
149
138
150
**Why this matters:** Catching grounding failures at the planning stage is faster than rewriting generated code. If the plan reveals shallow grounding, add constraints and force deeper research before execution.
Modern terminal options worth exploring: [**Ghostty**](https://ghostty.org) (fast, GPU-accelerated, native), [**Kitty**](https://sw.kovidgoyal.net/kitty/) (GPU-based, extensive graphics support), [**WezTerm**](https://wezterm.org) (Lua-configured, cross-platform), and [**Alacritty**](https://alacritty.org) (minimalist, OpenGL-accelerated). Each offers different customization approaches—compare based on your workflow needs.
183
195
184
-
**Use ArguSeek to learn terminal customization.** Research best practices for your chosen terminal, especially around session management, keybindings for rapid context switching, notification configuration, and visual indicators for different agent contexts. Example prompt: "Use ArguSeek to research Kitty terminal customization for managing multiple development sessions with different contexts and long-running processes."
196
+
**Use ArguSeek to learn terminal customization.** Research best practices for your chosen terminal, especially around session management, keybindings for rapid context switching, notification configuration, and visual indicators for different agent contexts. Example prompt:
197
+
198
+
```markdown
199
+
Use ArguSeek to research Kitty terminal customization for managing multiple
200
+
development sessions with different contexts and long-running processes.
201
+
```
185
202
186
203
### Modern CLI Tools for Efficient Workflows
187
204
@@ -217,7 +234,7 @@ The agent will research worktree workflows, propose a clean directory layout, an
217
234
218
235
### Mix CLI and UI Tools: Use What Works
219
236
220
-
Don't be dogmatic about terminal-only or GUI-only workflows. IDEs remain the best tools for code navigation, symbol search, and refactoring. CLI excels at quick edits, git operations, and managing parallel sessions.
237
+
Don't be dogmatic about terminal-only or GUI-only workflows. IDEs remain the best tools for code navigation, symbol search, and viewing large files. CLI excels at quick edits, git operations, and managing parallel sessions.
0 commit comments