@ckelsoe/prompt-architect 3.0.2 → 3.1.1

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
Files changed (76) hide show
  1. package/.claude-plugin/marketplace.json +1 -1
  2. package/.claude-plugin/plugin.json +1 -1
  3. package/CHANGELOG.md +264 -264
  4. package/LICENSE +21 -21
  5. package/README.md +1064 -1184
  6. package/adapters/README.md +75 -267
  7. package/adapters/for-windsurf.md +136 -136
  8. package/adapters/system-prompt.md +243 -243
  9. package/package.json +139 -138
  10. package/scripts/build-skill.js +64 -0
  11. package/scripts/install.js +4 -54
  12. package/scripts/test.js +205 -213
  13. package/scripts/validate-skill.js +350 -350
  14. package/skills/prompt-architect/SKILL.md +331 -331
  15. package/skills/prompt-architect/assets/templates/ape_template.txt +5 -5
  16. package/skills/prompt-architect/assets/templates/bab_template.txt +15 -15
  17. package/skills/prompt-architect/assets/templates/broke_template.txt +15 -15
  18. package/skills/prompt-architect/assets/templates/cai-critique-revise_template.txt +18 -18
  19. package/skills/prompt-architect/assets/templates/care_template.txt +16 -16
  20. package/skills/prompt-architect/assets/templates/chain-of-density_template.txt +46 -46
  21. package/skills/prompt-architect/assets/templates/chain-of-thought_template.txt +49 -49
  22. package/skills/prompt-architect/assets/templates/co-star_template.txt +17 -17
  23. package/skills/prompt-architect/assets/templates/crispe_template.txt +14 -14
  24. package/skills/prompt-architect/assets/templates/ctf_template.txt +15 -15
  25. package/skills/prompt-architect/assets/templates/devils-advocate_template.txt +25 -25
  26. package/skills/prompt-architect/assets/templates/hybrid_template.txt +101 -101
  27. package/skills/prompt-architect/assets/templates/least-to-most_template.txt +16 -16
  28. package/skills/prompt-architect/assets/templates/plan-and-solve_template.txt +7 -7
  29. package/skills/prompt-architect/assets/templates/pre-mortem_template.txt +27 -27
  30. package/skills/prompt-architect/assets/templates/race_template.txt +19 -19
  31. package/skills/prompt-architect/assets/templates/rcot_template.txt +31 -31
  32. package/skills/prompt-architect/assets/templates/react_template.txt +27 -27
  33. package/skills/prompt-architect/assets/templates/reverse-role_template.txt +11 -11
  34. package/skills/prompt-architect/assets/templates/rise-ie_template.txt +23 -23
  35. package/skills/prompt-architect/assets/templates/rise-ix_template.txt +28 -28
  36. package/skills/prompt-architect/assets/templates/risen_template.txt +21 -21
  37. package/skills/prompt-architect/assets/templates/rpef_template.txt +26 -26
  38. package/skills/prompt-architect/assets/templates/rtf_template.txt +13 -13
  39. package/skills/prompt-architect/assets/templates/self-refine_template.txt +16 -16
  40. package/skills/prompt-architect/assets/templates/skeleton-of-thought_template.txt +15 -15
  41. package/skills/prompt-architect/assets/templates/step-back_template.txt +10 -10
  42. package/skills/prompt-architect/assets/templates/tidd-ec_template.txt +45 -45
  43. package/skills/prompt-architect/assets/templates/tree-of-thought_template.txt +30 -30
  44. package/skills/prompt-architect/references/frameworks/ape.md +200 -200
  45. package/skills/prompt-architect/references/frameworks/bab.md +242 -242
  46. package/skills/prompt-architect/references/frameworks/broke.md +242 -242
  47. package/skills/prompt-architect/references/frameworks/cai-critique-revise.md +230 -230
  48. package/skills/prompt-architect/references/frameworks/care.md +235 -235
  49. package/skills/prompt-architect/references/frameworks/chain-of-density.md +479 -479
  50. package/skills/prompt-architect/references/frameworks/chain-of-thought.md +424 -424
  51. package/skills/prompt-architect/references/frameworks/co-star.md +256 -256
  52. package/skills/prompt-architect/references/frameworks/crispe.md +253 -253
  53. package/skills/prompt-architect/references/frameworks/ctf.md +207 -207
  54. package/skills/prompt-architect/references/frameworks/devils-advocate.md +206 -206
  55. package/skills/prompt-architect/references/frameworks/least-to-most.md +226 -226
  56. package/skills/prompt-architect/references/frameworks/plan-and-solve.md +210 -210
  57. package/skills/prompt-architect/references/frameworks/pre-mortem.md +197 -197
  58. package/skills/prompt-architect/references/frameworks/race.md +243 -243
  59. package/skills/prompt-architect/references/frameworks/rcot.md +203 -203
  60. package/skills/prompt-architect/references/frameworks/react.md +304 -304
  61. package/skills/prompt-architect/references/frameworks/reverse-role.md +189 -189
  62. package/skills/prompt-architect/references/frameworks/rise.md +555 -555
  63. package/skills/prompt-architect/references/frameworks/risen.md +297 -297
  64. package/skills/prompt-architect/references/frameworks/rpef.md +195 -195
  65. package/skills/prompt-architect/references/frameworks/rtf.md +358 -358
  66. package/skills/prompt-architect/references/frameworks/self-refine.md +205 -205
  67. package/skills/prompt-architect/references/frameworks/skeleton-of-thought.md +186 -186
  68. package/skills/prompt-architect/references/frameworks/step-back.md +206 -206
  69. package/skills/prompt-architect/references/frameworks/tidd-ec.md +470 -470
  70. package/skills/prompt-architect/references/frameworks/tree-of-thought.md +260 -260
  71. package/skills/prompt-architect/scripts/framework_analyzer.py +807 -807
  72. package/skills/prompt-architect/scripts/prompt_evaluator.py +336 -336
  73. package/adapters/for-cursor.mdc +0 -138
  74. package/adapters/for-gemini-cli.md +0 -70
  75. package/adapters/for-github-copilot.md +0 -141
  76. package/adapters/for-openai-codex-cli.md +0 -158
@@ -1,226 +1,226 @@
1
- # Least-to-Most (LtM) Prompting
2
-
3
- ## Overview
4
-
5
- Least-to-Most (LtM) prompting is a decomposition-first framework that breaks a complex problem into an ordered sequence of simpler subproblems, then solves them sequentially — using each prior answer as context for the next. Unlike Chain of Thought (which reasons through a problem in one pass), LtM explicitly separates decomposition from execution, and the subproblems are sequenced from simplest to most complex.
6
-
7
- **Research basis:** "Least-to-Most Prompting Enables Complex Reasoning in Large Language Models" (Zhou et al., Google Brain, arXiv:2205.10625, ICLR 2023). On SCAN compositional generalization: LtM achieves 99.7% vs. CoT's 16%. On GSM8K math: matches or exceeds CoT with fewer errors.
8
-
9
- ## Why LtM Differs from Chain of Thought
10
-
11
- | | Chain of Thought | Least-to-Most |
12
- |---|---|---|
13
- | Structure | Single-pass linear reasoning | Explicit decompose → solve sequence |
14
- | Decomposition | Implicit (within the chain) | Explicit first step |
15
- | Subproblem order | Not explicitly structured | Simplest → most complex |
16
- | Answer reuse | May or may not reference prior steps | Each answer feeds explicitly into the next |
17
- | Best for | Known procedure, moderate complexity | Compositional tasks, building-block problems |
18
-
19
- ## Components
20
-
21
- ### Phase 1: Decomposition
22
- **Purpose:** Break the original problem into an ordered list of subproblems, from simplest to most complex. The key: later subproblems should build on earlier ones.
23
-
24
- **Trigger:** "To solve [problem], we need to first solve: [list subproblems in order]"
25
-
26
- **Questions to Ask:**
27
- - What is the simplest thing that must be true before we can answer the full question?
28
- - What building blocks does the final answer depend on?
29
- - What is the correct ordering — which must come first?
30
-
31
- ### Phase 2: Sequential Solving
32
- **Purpose:** Solve each subproblem in order, explicitly using prior answers as context.
33
-
34
- **Pattern:** Answer subproblem 1 → use that answer when solving subproblem 2 → use both when solving subproblem 3 → ... → solve original problem.
35
-
36
- ## Template Structure
37
-
38
- ```
39
- PROBLEM:
40
- [The full, complex problem to solve]
41
-
42
- DECOMPOSE:
43
- Break this problem into ordered subproblems, from simplest to most complex.
44
- Each later subproblem should build on the answer to the previous.
45
-
46
- Subproblem 1: [Simplest prerequisite]
47
- Subproblem 2: [Next level, may depend on #1]
48
- Subproblem 3: [More complex, depends on #1-2]
49
- ...
50
- Final: [The original problem, now solvable using all prior answers]
51
-
52
- SOLVE SEQUENTIALLY:
53
- Solve each subproblem in order. Use prior answers as context for each
54
- subsequent subproblem. Show each solution before moving to the next.
55
- ```
56
-
57
- ### Single-Pass Trigger (model decomposes and solves itself):
58
- ```
59
- To solve the following problem, first identify the simpler subproblems
60
- that must be solved first (from least to most complex). Then solve each
61
- in sequence, using prior answers as context. Show decomposition, then
62
- sequential solutions.
63
-
64
- Problem: [your problem]
65
- ```
66
-
67
- ## Complete Examples
68
-
69
- ### Example 1: Multi-Hop Reasoning
70
-
71
- **Before LtM:**
72
- "What would the tax implications be for a US remote employee working from Portugal for 3 months?"
73
-
74
- **After LtM:**
75
- ```
76
- PROBLEM:
77
- What are the tax implications for a US citizen employed by a US company who
78
- works remotely from Portugal for 3 months in one calendar year?
79
-
80
- DECOMPOSE:
81
- Subproblem 1: What are the US tax rules for citizens earning income abroad
82
- (general principles, FEIE, FTC)?
83
-
84
- Subproblem 2: What is Portugal's tax rule for foreigners working there —
85
- specifically the threshold for establishing tax residency?
86
-
87
- Subproblem 3: Does 3 months in Portugal trigger Portuguese tax residency?
88
- (Use answers from 1 and 2)
89
-
90
- Subproblem 4: If both US and Portuguese tax obligations potentially apply,
91
- what does the US-Portugal tax treaty say about double taxation?
92
-
93
- Final: Given the above, what are the practical tax implications and what
94
- should the employee do?
95
-
96
- SOLVE SEQUENTIALLY:
97
- [Solve in order, each answer informing the next]
98
- ```
99
-
100
- ### Example 2: Code Architecture
101
-
102
- **Before LtM:**
103
- "How do I add real-time notifications to my Django app?"
104
-
105
- **After LtM:**
106
- ```
107
- PROBLEM:
108
- I need to add real-time push notifications to a Django app with PostgreSQL.
109
- Users should receive browser notifications when certain database records change.
110
-
111
- DECOMPOSE:
112
- Subproblem 1: What are the available mechanisms for real-time browser
113
- communication in Django? (WebSockets, SSE, polling — trade-offs)
114
-
115
- Subproblem 2: For the chosen mechanism, what Django-specific libraries
116
- and setup are required?
117
-
118
- Subproblem 3: How do we detect database record changes and trigger
119
- notifications? (Django signals, PostgreSQL LISTEN/NOTIFY, or polling)
120
-
121
- Subproblem 4: How do we route a database event → notification trigger →
122
- connected browser client?
123
-
124
- Final: What is the complete implementation plan and any gotchas?
125
-
126
- SOLVE SEQUENTIALLY:
127
- [Solve in order]
128
- ```
129
-
130
- ### Example 3: Business Analysis
131
-
132
- **Before LtM:**
133
- "Should we expand to the European market?"
134
-
135
- **After LtM:**
136
- ```
137
- PROBLEM:
138
- Our B2B SaaS has strong US traction. Should we expand to the European market
139
- now, given our current 20-person team and 18 months of runway?
140
-
141
- DECOMPOSE:
142
- Subproblem 1: What are the key differences between US and European B2B SaaS
143
- markets that affect a US-native product? (GDPR, sales cycles, localization needs)
144
-
145
- Subproblem 2: What resources are typically required for a credible EU market entry?
146
- (team, timeline, cost estimates)
147
-
148
- Subproblem 3: What are the signals that a company is or isn't ready for
149
- international expansion?
150
-
151
- Subproblem 4: Given our specific constraints (20 people, 18 months runway),
152
- how do our resources map to those requirements?
153
-
154
- Final: Should we expand to Europe now, wait, or explore a different
155
- expansion approach? With clear reasoning from the above.
156
-
157
- SOLVE SEQUENTIALLY:
158
- [Solve each in order, carry answers forward]
159
- ```
160
-
161
- ## Best Use Cases
162
-
163
- 1. **Compositional Reasoning**
164
- - Multi-hop questions requiring chained facts
165
- - Problems where the answer depends on several intermediate results
166
- - Any question with "it depends on X, which depends on Y"
167
-
168
- 2. **Complex Technical Problems**
169
- - Architecture decisions with prerequisites
170
- - Debugging with unknown root cause
171
- - Implementation planning with dependencies
172
-
173
- 3. **Multi-Domain Questions**
174
- - Questions spanning legal + technical, financial + operational, etc.
175
- - Any question where different areas of knowledge must be combined
176
-
177
- 4. **Long Math / Calculation Chains**
178
- - Financial projections with multiple steps
179
- - Physics problems with multiple formulas
180
- - Anything where intermediate errors cascade
181
-
182
- 5. **When Chain of Thought Keeps Failing**
183
- - LtM's explicit decomposition reduces error propagation
184
- - If CoT gives wrong answers, decomposing first often helps
185
-
186
- ## Selection Criteria
187
-
188
- **Choose LtM when:**
189
- - ✅ Problem has multiple prerequisite subproblems
190
- - ✅ Subproblems have a natural ordering (simplest first)
191
- - ✅ Each answer genuinely builds on the previous
192
- - ✅ CoT gives shallow or wrong answers on this type of question
193
- - ✅ Decomposition is non-trivial (question is compositional)
194
-
195
- **Avoid LtM when:**
196
- - ❌ Problem is simple and self-contained → use CoT or APE
197
- - ❌ No natural decomposition exists → use Tree of Thought instead
198
- - ❌ Subproblems are independent (no ordering needed) → use SoT or ToT
199
- - ❌ Content creation task → use CO-STAR or BAB
200
-
201
- ## Common Mistakes
202
-
203
- 1. **Subproblems That Don't Build**
204
- - Each subproblem should be a prerequisite for the next
205
- - If they're independent, that's SoT or Tree of Thought, not LtM
206
-
207
- 2. **Wrong Ordering**
208
- - More complex ≠ later; it's about dependency, not complexity per se
209
- - "What is X?" must come before "Given X, what is Y?"
210
-
211
- 3. **Not Carrying Answers Forward**
212
- - The key to LtM is that each answer is explicitly used in the next step
213
- - Don't solve subproblems in isolation
214
-
215
- 4. **Using LtM When CoT Is Sufficient**
216
- - If the problem has one reasoning path, CoT is simpler
217
- - LtM overhead is only worth it for genuinely compositional problems
218
-
219
- ## Quick Reference
220
-
221
- | Phase | Purpose | Key Question |
222
- |-------|---------|--------------|
223
- | Decompose | Order the subproblems | "What must I know first?" |
224
- | Solve (simplest) | Build the foundation | "What's the answer to this piece?" |
225
- | Solve (sequentially) | Build on prior answers | "Given what we know, what's next?" |
226
- | Final solve | Assemble the answer | "Now that I have all pieces, what's the answer?" |
1
+ # Least-to-Most (LtM) Prompting
2
+
3
+ ## Overview
4
+
5
+ Least-to-Most (LtM) prompting is a decomposition-first framework that breaks a complex problem into an ordered sequence of simpler subproblems, then solves them sequentially — using each prior answer as context for the next. Unlike Chain of Thought (which reasons through a problem in one pass), LtM explicitly separates decomposition from execution, and the subproblems are sequenced from simplest to most complex.
6
+
7
+ **Research basis:** "Least-to-Most Prompting Enables Complex Reasoning in Large Language Models" (Zhou et al., Google Brain, arXiv:2205.10625, ICLR 2023). On SCAN compositional generalization: LtM achieves 99.7% vs. CoT's 16%. On GSM8K math: matches or exceeds CoT with fewer errors.
8
+
9
+ ## Why LtM Differs from Chain of Thought
10
+
11
+ | | Chain of Thought | Least-to-Most |
12
+ |---|---|---|
13
+ | Structure | Single-pass linear reasoning | Explicit decompose → solve sequence |
14
+ | Decomposition | Implicit (within the chain) | Explicit first step |
15
+ | Subproblem order | Not explicitly structured | Simplest → most complex |
16
+ | Answer reuse | May or may not reference prior steps | Each answer feeds explicitly into the next |
17
+ | Best for | Known procedure, moderate complexity | Compositional tasks, building-block problems |
18
+
19
+ ## Components
20
+
21
+ ### Phase 1: Decomposition
22
+ **Purpose:** Break the original problem into an ordered list of subproblems, from simplest to most complex. The key: later subproblems should build on earlier ones.
23
+
24
+ **Trigger:** "To solve [problem], we need to first solve: [list subproblems in order]"
25
+
26
+ **Questions to Ask:**
27
+ - What is the simplest thing that must be true before we can answer the full question?
28
+ - What building blocks does the final answer depend on?
29
+ - What is the correct ordering — which must come first?
30
+
31
+ ### Phase 2: Sequential Solving
32
+ **Purpose:** Solve each subproblem in order, explicitly using prior answers as context.
33
+
34
+ **Pattern:** Answer subproblem 1 → use that answer when solving subproblem 2 → use both when solving subproblem 3 → ... → solve original problem.
35
+
36
+ ## Template Structure
37
+
38
+ ```
39
+ PROBLEM:
40
+ [The full, complex problem to solve]
41
+
42
+ DECOMPOSE:
43
+ Break this problem into ordered subproblems, from simplest to most complex.
44
+ Each later subproblem should build on the answer to the previous.
45
+
46
+ Subproblem 1: [Simplest prerequisite]
47
+ Subproblem 2: [Next level, may depend on #1]
48
+ Subproblem 3: [More complex, depends on #1-2]
49
+ ...
50
+ Final: [The original problem, now solvable using all prior answers]
51
+
52
+ SOLVE SEQUENTIALLY:
53
+ Solve each subproblem in order. Use prior answers as context for each
54
+ subsequent subproblem. Show each solution before moving to the next.
55
+ ```
56
+
57
+ ### Single-Pass Trigger (model decomposes and solves itself):
58
+ ```
59
+ To solve the following problem, first identify the simpler subproblems
60
+ that must be solved first (from least to most complex). Then solve each
61
+ in sequence, using prior answers as context. Show decomposition, then
62
+ sequential solutions.
63
+
64
+ Problem: [your problem]
65
+ ```
66
+
67
+ ## Complete Examples
68
+
69
+ ### Example 1: Multi-Hop Reasoning
70
+
71
+ **Before LtM:**
72
+ "What would the tax implications be for a US remote employee working from Portugal for 3 months?"
73
+
74
+ **After LtM:**
75
+ ```
76
+ PROBLEM:
77
+ What are the tax implications for a US citizen employed by a US company who
78
+ works remotely from Portugal for 3 months in one calendar year?
79
+
80
+ DECOMPOSE:
81
+ Subproblem 1: What are the US tax rules for citizens earning income abroad
82
+ (general principles, FEIE, FTC)?
83
+
84
+ Subproblem 2: What is Portugal's tax rule for foreigners working there —
85
+ specifically the threshold for establishing tax residency?
86
+
87
+ Subproblem 3: Does 3 months in Portugal trigger Portuguese tax residency?
88
+ (Use answers from 1 and 2)
89
+
90
+ Subproblem 4: If both US and Portuguese tax obligations potentially apply,
91
+ what does the US-Portugal tax treaty say about double taxation?
92
+
93
+ Final: Given the above, what are the practical tax implications and what
94
+ should the employee do?
95
+
96
+ SOLVE SEQUENTIALLY:
97
+ [Solve in order, each answer informing the next]
98
+ ```
99
+
100
+ ### Example 2: Code Architecture
101
+
102
+ **Before LtM:**
103
+ "How do I add real-time notifications to my Django app?"
104
+
105
+ **After LtM:**
106
+ ```
107
+ PROBLEM:
108
+ I need to add real-time push notifications to a Django app with PostgreSQL.
109
+ Users should receive browser notifications when certain database records change.
110
+
111
+ DECOMPOSE:
112
+ Subproblem 1: What are the available mechanisms for real-time browser
113
+ communication in Django? (WebSockets, SSE, polling — trade-offs)
114
+
115
+ Subproblem 2: For the chosen mechanism, what Django-specific libraries
116
+ and setup are required?
117
+
118
+ Subproblem 3: How do we detect database record changes and trigger
119
+ notifications? (Django signals, PostgreSQL LISTEN/NOTIFY, or polling)
120
+
121
+ Subproblem 4: How do we route a database event → notification trigger →
122
+ connected browser client?
123
+
124
+ Final: What is the complete implementation plan and any gotchas?
125
+
126
+ SOLVE SEQUENTIALLY:
127
+ [Solve in order]
128
+ ```
129
+
130
+ ### Example 3: Business Analysis
131
+
132
+ **Before LtM:**
133
+ "Should we expand to the European market?"
134
+
135
+ **After LtM:**
136
+ ```
137
+ PROBLEM:
138
+ Our B2B SaaS has strong US traction. Should we expand to the European market
139
+ now, given our current 20-person team and 18 months of runway?
140
+
141
+ DECOMPOSE:
142
+ Subproblem 1: What are the key differences between US and European B2B SaaS
143
+ markets that affect a US-native product? (GDPR, sales cycles, localization needs)
144
+
145
+ Subproblem 2: What resources are typically required for a credible EU market entry?
146
+ (team, timeline, cost estimates)
147
+
148
+ Subproblem 3: What are the signals that a company is or isn't ready for
149
+ international expansion?
150
+
151
+ Subproblem 4: Given our specific constraints (20 people, 18 months runway),
152
+ how do our resources map to those requirements?
153
+
154
+ Final: Should we expand to Europe now, wait, or explore a different
155
+ expansion approach? With clear reasoning from the above.
156
+
157
+ SOLVE SEQUENTIALLY:
158
+ [Solve each in order, carry answers forward]
159
+ ```
160
+
161
+ ## Best Use Cases
162
+
163
+ 1. **Compositional Reasoning**
164
+ - Multi-hop questions requiring chained facts
165
+ - Problems where the answer depends on several intermediate results
166
+ - Any question with "it depends on X, which depends on Y"
167
+
168
+ 2. **Complex Technical Problems**
169
+ - Architecture decisions with prerequisites
170
+ - Debugging with unknown root cause
171
+ - Implementation planning with dependencies
172
+
173
+ 3. **Multi-Domain Questions**
174
+ - Questions spanning legal + technical, financial + operational, etc.
175
+ - Any question where different areas of knowledge must be combined
176
+
177
+ 4. **Long Math / Calculation Chains**
178
+ - Financial projections with multiple steps
179
+ - Physics problems with multiple formulas
180
+ - Anything where intermediate errors cascade
181
+
182
+ 5. **When Chain of Thought Keeps Failing**
183
+ - LtM's explicit decomposition reduces error propagation
184
+ - If CoT gives wrong answers, decomposing first often helps
185
+
186
+ ## Selection Criteria
187
+
188
+ **Choose LtM when:**
189
+ - ✅ Problem has multiple prerequisite subproblems
190
+ - ✅ Subproblems have a natural ordering (simplest first)
191
+ - ✅ Each answer genuinely builds on the previous
192
+ - ✅ CoT gives shallow or wrong answers on this type of question
193
+ - ✅ Decomposition is non-trivial (question is compositional)
194
+
195
+ **Avoid LtM when:**
196
+ - ❌ Problem is simple and self-contained → use CoT or APE
197
+ - ❌ No natural decomposition exists → use Tree of Thought instead
198
+ - ❌ Subproblems are independent (no ordering needed) → use SoT or ToT
199
+ - ❌ Content creation task → use CO-STAR or BAB
200
+
201
+ ## Common Mistakes
202
+
203
+ 1. **Subproblems That Don't Build**
204
+ - Each subproblem should be a prerequisite for the next
205
+ - If they're independent, that's SoT or Tree of Thought, not LtM
206
+
207
+ 2. **Wrong Ordering**
208
+ - More complex ≠ later; it's about dependency, not complexity per se
209
+ - "What is X?" must come before "Given X, what is Y?"
210
+
211
+ 3. **Not Carrying Answers Forward**
212
+ - The key to LtM is that each answer is explicitly used in the next step
213
+ - Don't solve subproblems in isolation
214
+
215
+ 4. **Using LtM When CoT Is Sufficient**
216
+ - If the problem has one reasoning path, CoT is simpler
217
+ - LtM overhead is only worth it for genuinely compositional problems
218
+
219
+ ## Quick Reference
220
+
221
+ | Phase | Purpose | Key Question |
222
+ |-------|---------|--------------|
223
+ | Decompose | Order the subproblems | "What must I know first?" |
224
+ | Solve (simplest) | Build the foundation | "What's the answer to this piece?" |
225
+ | Solve (sequentially) | Build on prior answers | "Given what we know, what's next?" |
226
+ | Final solve | Assemble the answer | "Now that I have all pieces, what's the answer?" |