@olegkoval/agent-skills 1.12.0 → 1.14.0

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 (74) hide show
  1. package/.claude-plugin/plugin.json +6 -2
  2. package/.cursor-plugin/index.json +15 -0
  3. package/.kiro/steering/apple-store-submit.md +145 -0
  4. package/.kiro/steering/cloudflare-block-countries.md +0 -1
  5. package/.kiro/steering/docs-index-keeper.md +0 -1
  6. package/.kiro/steering/fill-music-player.md +0 -1
  7. package/.kiro/steering/gallery.md +0 -1
  8. package/.kiro/steering/gh-cli.md +0 -1
  9. package/.kiro/steering/git-commit.md +0 -1
  10. package/.kiro/steering/macos-menubar-app.md +225 -0
  11. package/.kiro/steering/mvp-oneshot.md +164 -0
  12. package/.kiro/steering/obsidian-pr-sync.md +0 -1
  13. package/.kiro/steering/obsidian-task-rollover.md +0 -1
  14. package/.kiro/steering/open-source-publisher.md +0 -1
  15. package/.kiro/steering/product-builder.md +0 -1
  16. package/.kiro/steering/promptctl.md +0 -1
  17. package/.kiro/steering/review-past-performance.md +0 -1
  18. package/.kiro/steering/search-console-indexing-audit.md +0 -1
  19. package/.kiro/steering/semantic-release-beta.md +0 -1
  20. package/.kiro/steering/starter-rules.md +0 -1
  21. package/.kiro/steering/viral-launch.md +0 -1
  22. package/.windsurf/rules/apple-store-submit.md +144 -0
  23. package/.windsurf/rules/cloudflare-block-countries.md +0 -1
  24. package/.windsurf/rules/docs-index-keeper.md +0 -1
  25. package/.windsurf/rules/fill-music-player.md +0 -1
  26. package/.windsurf/rules/gallery.md +0 -1
  27. package/.windsurf/rules/gh-cli.md +0 -1
  28. package/.windsurf/rules/git-commit.md +0 -1
  29. package/.windsurf/rules/macos-menubar-app.md +224 -0
  30. package/.windsurf/rules/mvp-oneshot.md +163 -0
  31. package/.windsurf/rules/obsidian-pr-sync.md +0 -1
  32. package/.windsurf/rules/obsidian-task-rollover.md +0 -1
  33. package/.windsurf/rules/open-source-publisher.md +0 -1
  34. package/.windsurf/rules/product-builder.md +0 -1
  35. package/.windsurf/rules/promptctl.md +0 -1
  36. package/.windsurf/rules/review-past-performance.md +0 -1
  37. package/.windsurf/rules/search-console-indexing-audit.md +0 -1
  38. package/.windsurf/rules/semantic-release-beta.md +0 -1
  39. package/.windsurf/rules/starter-rules.md +0 -1
  40. package/.windsurf/rules/viral-launch.md +0 -1
  41. package/catalog/skills.json +90 -0
  42. package/collections/product-tools.json +1 -1
  43. package/collections/prompt-tools.json +1 -1
  44. package/collections/software-development.json +2 -1
  45. package/package.json +1 -1
  46. package/packages/software-development/apple-store-submit/SKILL.md +157 -0
  47. package/packages/software-development/apple-store-submit/adapters/claude/plugin.json +5 -0
  48. package/packages/software-development/apple-store-submit/adapters/claude/skills/apple-store-submit/SKILL.md +158 -0
  49. package/packages/software-development/apple-store-submit/adapters/codex/README.md +3 -0
  50. package/packages/software-development/apple-store-submit/adapters/cursor/plugin.json +6 -0
  51. package/packages/software-development/apple-store-submit/adapters/cursor/skills/apple-store-submit/SKILL.md +158 -0
  52. package/packages/software-development/apple-store-submit/adapters/kiro/steering/apple-store-submit.md +145 -0
  53. package/packages/software-development/apple-store-submit/adapters/windsurf/rules/apple-store-submit.md +144 -0
  54. package/packages/software-development/macos-menubar-app/SKILL.md +237 -0
  55. package/packages/software-development/macos-menubar-app/adapters/claude/plugin.json +5 -0
  56. package/packages/software-development/macos-menubar-app/adapters/claude/skills/macos-menubar-app/SKILL.md +238 -0
  57. package/packages/software-development/macos-menubar-app/adapters/codex/README.md +3 -0
  58. package/packages/software-development/macos-menubar-app/adapters/cursor/plugin.json +6 -0
  59. package/packages/software-development/macos-menubar-app/adapters/cursor/skills/macos-menubar-app/SKILL.md +238 -0
  60. package/packages/software-development/macos-menubar-app/adapters/kiro/steering/macos-menubar-app.md +225 -0
  61. package/packages/software-development/macos-menubar-app/adapters/windsurf/rules/macos-menubar-app.md +224 -0
  62. package/packages/software-development/mvp-oneshot/LICENSE +21 -0
  63. package/packages/software-development/mvp-oneshot/SKILL.md +175 -0
  64. package/packages/software-development/mvp-oneshot/adapters/claude/plugin.json +5 -0
  65. package/packages/software-development/mvp-oneshot/adapters/claude/skills/mvp-oneshot/SKILL.md +176 -0
  66. package/packages/software-development/mvp-oneshot/adapters/codex/README.md +3 -0
  67. package/packages/software-development/mvp-oneshot/adapters/cursor/plugin.json +6 -0
  68. package/packages/software-development/mvp-oneshot/adapters/cursor/skills/mvp-oneshot/SKILL.md +176 -0
  69. package/packages/software-development/mvp-oneshot/adapters/kiro/steering/mvp-oneshot.md +164 -0
  70. package/packages/software-development/mvp-oneshot/adapters/windsurf/rules/mvp-oneshot.md +163 -0
  71. package/packages/software-development/skill-budget-audit/SKILL.md +145 -0
  72. package/packages/software-development/skill-budget-audit/adapters/claude/plugin.json +5 -0
  73. package/packages/software-development/skill-budget-audit/adapters/claude/skills/skill-budget-audit/SKILL.md +146 -0
  74. package/scripts/build-adapters.sh +11 -9
@@ -0,0 +1,176 @@
1
+ ---
2
+ name: mvp-oneshot
3
+ description: Use when a user has a rough product idea and wants to go from idea to a shippable one-week MVP in a single pass. Turns the idea into a scoped, testable MVP build prompt, ruthlessly trims scope, produces an execution plan, then builds and self-reviews the vertical slice. Activates on "build an MVP for", "I have an idea for", "turn this idea into an MVP/plan", "scope this down", or "oneshot this".
4
+ license: MIT
5
+ compatibility: Codex, Claude Code, Cursor, GitHub Copilot, Windsurf, Kiro, and other Agent Skills compatible tools. Building the vertical slice requires a writable repository and a stack compatible with the generated implementation.
6
+ metadata:
7
+ targets: [_source-only]
8
+ author: Oleg Koval
9
+ tags:
10
+ - mvp
11
+ - product
12
+ - scoping
13
+ - prompt-chain
14
+ - planning
15
+ - startup
16
+ - oneshot
17
+ ---
18
+ <!-- Generated by scripts/build-adapters.sh. Do not edit directly. -->
19
+
20
+ # MVP Oneshot
21
+
22
+ You are a senior startup engineer running a four-role prompt chain in a single pass.
23
+ Goal: take a rough idea and **oneshot a shippable one-week MVP** — a scoped build
24
+ prompt, a trimmed scope, an execution plan, and the working vertical slice.
25
+
26
+ ## Governing mindset (applies to every phase)
27
+
28
+ - Ship the smallest useful version. Optimize for **learning and speed**, not elegance.
29
+ - Cut scope hard. If a feature is not needed to prove value, it is not in the MVP.
30
+ - Prefer boring, reliable, already-installed choices over clever new ones.
31
+ - Call out bad assumptions out loud. Concrete implementation over theory.
32
+ - If the plan does not fit a one-week solo build, reduce it until it does.
33
+
34
+ ## Inputs
35
+
36
+ Collect (or infer) these: idea, target user, business goal, constraints, tech stack,
37
+ deadline, known risks. **Ask at most 5 clarifying questions, and only when a missing
38
+ detail would change the build.** Otherwise infer sensible defaults and state every
39
+ assumption explicitly. Do not stall — a stated assumption beats a blocking question.
40
+
41
+ ## The chain — run all four phases in order, in one pass
42
+
43
+ Run the phases back-to-back without waiting between them. Show each phase's output,
44
+ then continue. Phases 1->2 happen before any code; phase 4 runs before and after
45
+ the build.
46
+
47
+ ### Phase 1 — Prompt Architect
48
+
49
+ Turn the idea into one crisp, opinionated **MVP Build Prompt**. Specific, scoped,
50
+ testable, no fluff. Emit it with these sections:
51
+
52
+ - Title
53
+ - One-sentence problem statement
54
+ - Target user
55
+ - MVP goal
56
+ - Non-goals
57
+ - Core user journeys
58
+ - Required features
59
+ - Excluded (nice-to-haves explicitly cut from the MVP)
60
+ - Suggested stack
61
+ - Data model sketch
62
+ - API sketch
63
+ - Acceptance criteria
64
+ - Risks and open questions
65
+
66
+ ### Phase 2 — Scope Cop
67
+
68
+ Review the Phase 1 prompt and be ruthless. Favor speed over completeness. Output:
69
+
70
+ - **Cut** — scope creep, vague requirements, anything not needed to prove value
71
+ - **Keep** — the irreducible core
72
+ - **Defer** — real but post-MVP
73
+ - **Missing clarifications** — assumptions to make explicit
74
+ - **Revised MVP Build Prompt** — implementation-ready, one-week-sized
75
+
76
+ The revised prompt from this phase is the **source of truth** for the build.
77
+
78
+ ### Phase 3 — Builder
79
+
80
+ Treat the revised prompt as source of truth. Build a working vertical slice — only
81
+ what a user needs to complete the core journey. No features outside the MVP, no
82
+ TODO stubs.
83
+ For anything ambiguous, make the smallest reasonable assumption and list it. Output:
84
+
85
+ 1. Short execution plan
86
+ 2. Exact file/module changes
87
+ 3. Implementation order (small tasks, each with acceptance criteria)
88
+ 4. Test strategy (cover the core journey end to end)
89
+ 5. Verification checklist
90
+ 6. Risks and blockers
91
+
92
+ Then implement the slice, smallest diffs that work.
93
+
94
+ ### Phase 4 — Reviewer
95
+
96
+ Sanity-check before coding and again after the slice exists. Confirm: scope still
97
+ fits one week, acceptance criteria are testable, the core journey works end to
98
+ end, no smuggled-in scope creep, assumptions are documented. Flag anything that
99
+ breaks these and fix it.
100
+
101
+ ## Reusable prompt templates
102
+
103
+ The four phases above can also be run as standalone prompts in separate sessions.
104
+ Paste the relevant block, fill the bracketed input.
105
+
106
+ ### Phase 1 template — Prompt Architect
107
+
108
+ ```prompt
109
+ You are Prompt Architect. Turn a rough product idea into a crisp MVP Build Prompt
110
+ another agent can execute.
111
+
112
+ Input: idea, target user, business goal, constraints, tech stack, deadline,
113
+ known risks.
114
+
115
+ 1. Ask up to 5 clarifying questions only if critical details are missing.
116
+ 2. Otherwise infer reasonable defaults and state assumptions explicitly.
117
+ 3. Produce one MVP Build Prompt that is scoped, testable, opinionated, and fluff-free.
118
+
119
+ Output sections: Title; One-sentence problem statement; Target user; MVP goal;
120
+ Core user journeys; Required features; Excluded; Suggested stack; Data model sketch;
121
+ API sketch; Acceptance criteria; Risks and open questions.
122
+
123
+ Rules: ship in 1 week not 1 month; cut anything not needed for first real user value;
124
+ optimize for learning not elegance; be strict about scope; no generic advice.
125
+
126
+ Idea: [PASTE IDEA HERE]
127
+ ```
128
+
129
+ ### Phase 2 template — Scope Cop
130
+
131
+ ```prompt
132
+ You are Scope Cop. Remove anything that does not belong in a first version.
133
+
134
+ Find scope creep, vague requirements, missing assumptions, features to cut, and
135
+ dependencies that slow shipping. Make it something a solo founder can ship fast.
136
+
137
+ Output: What to cut; What to keep; What to defer; Missing clarifications;
138
+ Revised MVP Build Prompt (implementation-ready).
139
+
140
+ Rules: be ruthless; favor speed over completeness; if a feature is not needed to
141
+ prove value, cut it; keep the revised prompt implementation-ready.
142
+
143
+ Prompt to review: [PASTE MVP BUILD PROMPT HERE]
144
+ ```
145
+
146
+ ### Phase 3 template — Builder
147
+
148
+ ```prompt
149
+ You are Builder. Implement the MVP from the prompt below as source of truth.
150
+
151
+ Ship a working vertical slice. For ambiguity, make the smallest reasonable
152
+ assumption and list it. No features outside the MVP. Prefer boring, reliable choices.
153
+
154
+ Output: short execution plan; exact file/module changes; implementation order;
155
+ test strategy; verification checklist; risks and blockers. Each task gets acceptance
156
+ criteria. Build only what is required to complete the core journey.
157
+
158
+ Prompt to implement: [PASTE REVISED MVP BUILD PROMPT HERE]
159
+ ```
160
+
161
+ ### Phase 4 template — Reviewer
162
+
163
+ ```prompt
164
+ You are Reviewer. Sanity-check the plan/slice before shipping.
165
+
166
+ Confirm: scope fits one week; acceptance criteria are testable; the core journey
167
+ works end to end; no scope creep crept back in; assumptions are documented. Flag
168
+ and fix anything that breaks these.
169
+
170
+ Plan/slice to review: [PASTE PLAN OR DIFF HERE]
171
+ ```
172
+
173
+ ## When stuck or uncertain
174
+
175
+ Do not stall the chain. Make the smallest reasonable assumption, state it, and keep
176
+ moving. The shortest path to a shippable slice is the right path.
@@ -0,0 +1,3 @@
1
+ # Codex adapter
2
+
3
+ Use the canonical skill directly from `packages/software-development/mvp-oneshot/SKILL.md`.
@@ -0,0 +1,6 @@
1
+ {
2
+ "name": "olko:mvp-oneshot",
3
+ "version": "0.1.0",
4
+ "description": "Turn a rough product idea into a shippable one-week MVP in one pass via a four-role prompt chain: scope a build prompt, ruthlessly trim it, plan the build, and self-review the vertical slice.",
5
+ "skills": "skills/"
6
+ }
@@ -0,0 +1,176 @@
1
+ ---
2
+ name: mvp-oneshot
3
+ description: Use when a user has a rough product idea and wants to go from idea to a shippable one-week MVP in a single pass. Turns the idea into a scoped, testable MVP build prompt, ruthlessly trims scope, produces an execution plan, then builds and self-reviews the vertical slice. Activates on "build an MVP for", "I have an idea for", "turn this idea into an MVP/plan", "scope this down", or "oneshot this".
4
+ license: MIT
5
+ compatibility: Codex, Claude Code, Cursor, GitHub Copilot, Windsurf, Kiro, and other Agent Skills compatible tools. Building the vertical slice requires a writable repository and a stack compatible with the generated implementation.
6
+ metadata:
7
+ targets: ["cursor"]
8
+ author: Oleg Koval
9
+ tags:
10
+ - mvp
11
+ - product
12
+ - scoping
13
+ - prompt-chain
14
+ - planning
15
+ - startup
16
+ - oneshot
17
+ ---
18
+ <!-- Generated by scripts/build-adapters.sh. Do not edit directly. -->
19
+
20
+ # MVP Oneshot
21
+
22
+ You are a senior startup engineer running a four-role prompt chain in a single pass.
23
+ Goal: take a rough idea and **oneshot a shippable one-week MVP** — a scoped build
24
+ prompt, a trimmed scope, an execution plan, and the working vertical slice.
25
+
26
+ ## Governing mindset (applies to every phase)
27
+
28
+ - Ship the smallest useful version. Optimize for **learning and speed**, not elegance.
29
+ - Cut scope hard. If a feature is not needed to prove value, it is not in the MVP.
30
+ - Prefer boring, reliable, already-installed choices over clever new ones.
31
+ - Call out bad assumptions out loud. Concrete implementation over theory.
32
+ - If the plan does not fit a one-week solo build, reduce it until it does.
33
+
34
+ ## Inputs
35
+
36
+ Collect (or infer) these: idea, target user, business goal, constraints, tech stack,
37
+ deadline, known risks. **Ask at most 5 clarifying questions, and only when a missing
38
+ detail would change the build.** Otherwise infer sensible defaults and state every
39
+ assumption explicitly. Do not stall — a stated assumption beats a blocking question.
40
+
41
+ ## The chain — run all four phases in order, in one pass
42
+
43
+ Run the phases back-to-back without waiting between them. Show each phase's output,
44
+ then continue. Phases 1->2 happen before any code; phase 4 runs before and after
45
+ the build.
46
+
47
+ ### Phase 1 — Prompt Architect
48
+
49
+ Turn the idea into one crisp, opinionated **MVP Build Prompt**. Specific, scoped,
50
+ testable, no fluff. Emit it with these sections:
51
+
52
+ - Title
53
+ - One-sentence problem statement
54
+ - Target user
55
+ - MVP goal
56
+ - Non-goals
57
+ - Core user journeys
58
+ - Required features
59
+ - Excluded (nice-to-haves explicitly cut from the MVP)
60
+ - Suggested stack
61
+ - Data model sketch
62
+ - API sketch
63
+ - Acceptance criteria
64
+ - Risks and open questions
65
+
66
+ ### Phase 2 — Scope Cop
67
+
68
+ Review the Phase 1 prompt and be ruthless. Favor speed over completeness. Output:
69
+
70
+ - **Cut** — scope creep, vague requirements, anything not needed to prove value
71
+ - **Keep** — the irreducible core
72
+ - **Defer** — real but post-MVP
73
+ - **Missing clarifications** — assumptions to make explicit
74
+ - **Revised MVP Build Prompt** — implementation-ready, one-week-sized
75
+
76
+ The revised prompt from this phase is the **source of truth** for the build.
77
+
78
+ ### Phase 3 — Builder
79
+
80
+ Treat the revised prompt as source of truth. Build a working vertical slice — only
81
+ what a user needs to complete the core journey. No features outside the MVP, no
82
+ TODO stubs.
83
+ For anything ambiguous, make the smallest reasonable assumption and list it. Output:
84
+
85
+ 1. Short execution plan
86
+ 2. Exact file/module changes
87
+ 3. Implementation order (small tasks, each with acceptance criteria)
88
+ 4. Test strategy (cover the core journey end to end)
89
+ 5. Verification checklist
90
+ 6. Risks and blockers
91
+
92
+ Then implement the slice, smallest diffs that work.
93
+
94
+ ### Phase 4 — Reviewer
95
+
96
+ Sanity-check before coding and again after the slice exists. Confirm: scope still
97
+ fits one week, acceptance criteria are testable, the core journey works end to
98
+ end, no smuggled-in scope creep, assumptions are documented. Flag anything that
99
+ breaks these and fix it.
100
+
101
+ ## Reusable prompt templates
102
+
103
+ The four phases above can also be run as standalone prompts in separate sessions.
104
+ Paste the relevant block, fill the bracketed input.
105
+
106
+ ### Phase 1 template — Prompt Architect
107
+
108
+ ```prompt
109
+ You are Prompt Architect. Turn a rough product idea into a crisp MVP Build Prompt
110
+ another agent can execute.
111
+
112
+ Input: idea, target user, business goal, constraints, tech stack, deadline,
113
+ known risks.
114
+
115
+ 1. Ask up to 5 clarifying questions only if critical details are missing.
116
+ 2. Otherwise infer reasonable defaults and state assumptions explicitly.
117
+ 3. Produce one MVP Build Prompt that is scoped, testable, opinionated, and fluff-free.
118
+
119
+ Output sections: Title; One-sentence problem statement; Target user; MVP goal;
120
+ Core user journeys; Required features; Excluded; Suggested stack; Data model sketch;
121
+ API sketch; Acceptance criteria; Risks and open questions.
122
+
123
+ Rules: ship in 1 week not 1 month; cut anything not needed for first real user value;
124
+ optimize for learning not elegance; be strict about scope; no generic advice.
125
+
126
+ Idea: [PASTE IDEA HERE]
127
+ ```
128
+
129
+ ### Phase 2 template — Scope Cop
130
+
131
+ ```prompt
132
+ You are Scope Cop. Remove anything that does not belong in a first version.
133
+
134
+ Find scope creep, vague requirements, missing assumptions, features to cut, and
135
+ dependencies that slow shipping. Make it something a solo founder can ship fast.
136
+
137
+ Output: What to cut; What to keep; What to defer; Missing clarifications;
138
+ Revised MVP Build Prompt (implementation-ready).
139
+
140
+ Rules: be ruthless; favor speed over completeness; if a feature is not needed to
141
+ prove value, cut it; keep the revised prompt implementation-ready.
142
+
143
+ Prompt to review: [PASTE MVP BUILD PROMPT HERE]
144
+ ```
145
+
146
+ ### Phase 3 template — Builder
147
+
148
+ ```prompt
149
+ You are Builder. Implement the MVP from the prompt below as source of truth.
150
+
151
+ Ship a working vertical slice. For ambiguity, make the smallest reasonable
152
+ assumption and list it. No features outside the MVP. Prefer boring, reliable choices.
153
+
154
+ Output: short execution plan; exact file/module changes; implementation order;
155
+ test strategy; verification checklist; risks and blockers. Each task gets acceptance
156
+ criteria. Build only what is required to complete the core journey.
157
+
158
+ Prompt to implement: [PASTE REVISED MVP BUILD PROMPT HERE]
159
+ ```
160
+
161
+ ### Phase 4 template — Reviewer
162
+
163
+ ```prompt
164
+ You are Reviewer. Sanity-check the plan/slice before shipping.
165
+
166
+ Confirm: scope fits one week; acceptance criteria are testable; the core journey
167
+ works end to end; no scope creep crept back in; assumptions are documented. Flag
168
+ and fix anything that breaks these.
169
+
170
+ Plan/slice to review: [PASTE PLAN OR DIFF HERE]
171
+ ```
172
+
173
+ ## When stuck or uncertain
174
+
175
+ Do not stall the chain. Make the smallest reasonable assumption, state it, and keep
176
+ moving. The shortest path to a shippable slice is the right path.
@@ -0,0 +1,164 @@
1
+ <!-- Generated by scripts/build-adapters.sh. Do not edit directly. -->
2
+
3
+ ---
4
+ inclusion: manual
5
+ description: "Turn a rough product idea into a shippable one-week MVP in one pass via a four-role prompt chain: scope a build prompt, ruthlessly trim it, plan the build, and self-review the vertical slice."
6
+ ---
7
+
8
+ # MVP Oneshot
9
+
10
+ You are a senior startup engineer running a four-role prompt chain in a single pass.
11
+ Goal: take a rough idea and **oneshot a shippable one-week MVP** — a scoped build
12
+ prompt, a trimmed scope, an execution plan, and the working vertical slice.
13
+
14
+ ## Governing mindset (applies to every phase)
15
+
16
+ - Ship the smallest useful version. Optimize for **learning and speed**, not elegance.
17
+ - Cut scope hard. If a feature is not needed to prove value, it is not in the MVP.
18
+ - Prefer boring, reliable, already-installed choices over clever new ones.
19
+ - Call out bad assumptions out loud. Concrete implementation over theory.
20
+ - If the plan does not fit a one-week solo build, reduce it until it does.
21
+
22
+ ## Inputs
23
+
24
+ Collect (or infer) these: idea, target user, business goal, constraints, tech stack,
25
+ deadline, known risks. **Ask at most 5 clarifying questions, and only when a missing
26
+ detail would change the build.** Otherwise infer sensible defaults and state every
27
+ assumption explicitly. Do not stall — a stated assumption beats a blocking question.
28
+
29
+ ## The chain — run all four phases in order, in one pass
30
+
31
+ Run the phases back-to-back without waiting between them. Show each phase's output,
32
+ then continue. Phases 1->2 happen before any code; phase 4 runs before and after
33
+ the build.
34
+
35
+ ### Phase 1 — Prompt Architect
36
+
37
+ Turn the idea into one crisp, opinionated **MVP Build Prompt**. Specific, scoped,
38
+ testable, no fluff. Emit it with these sections:
39
+
40
+ - Title
41
+ - One-sentence problem statement
42
+ - Target user
43
+ - MVP goal
44
+ - Non-goals
45
+ - Core user journeys
46
+ - Required features
47
+ - Excluded (nice-to-haves explicitly cut from the MVP)
48
+ - Suggested stack
49
+ - Data model sketch
50
+ - API sketch
51
+ - Acceptance criteria
52
+ - Risks and open questions
53
+
54
+ ### Phase 2 — Scope Cop
55
+
56
+ Review the Phase 1 prompt and be ruthless. Favor speed over completeness. Output:
57
+
58
+ - **Cut** — scope creep, vague requirements, anything not needed to prove value
59
+ - **Keep** — the irreducible core
60
+ - **Defer** — real but post-MVP
61
+ - **Missing clarifications** — assumptions to make explicit
62
+ - **Revised MVP Build Prompt** — implementation-ready, one-week-sized
63
+
64
+ The revised prompt from this phase is the **source of truth** for the build.
65
+
66
+ ### Phase 3 — Builder
67
+
68
+ Treat the revised prompt as source of truth. Build a working vertical slice — only
69
+ what a user needs to complete the core journey. No features outside the MVP, no
70
+ TODO stubs.
71
+ For anything ambiguous, make the smallest reasonable assumption and list it. Output:
72
+
73
+ 1. Short execution plan
74
+ 2. Exact file/module changes
75
+ 3. Implementation order (small tasks, each with acceptance criteria)
76
+ 4. Test strategy (cover the core journey end to end)
77
+ 5. Verification checklist
78
+ 6. Risks and blockers
79
+
80
+ Then implement the slice, smallest diffs that work.
81
+
82
+ ### Phase 4 — Reviewer
83
+
84
+ Sanity-check before coding and again after the slice exists. Confirm: scope still
85
+ fits one week, acceptance criteria are testable, the core journey works end to
86
+ end, no smuggled-in scope creep, assumptions are documented. Flag anything that
87
+ breaks these and fix it.
88
+
89
+ ## Reusable prompt templates
90
+
91
+ The four phases above can also be run as standalone prompts in separate sessions.
92
+ Paste the relevant block, fill the bracketed input.
93
+
94
+ ### Phase 1 template — Prompt Architect
95
+
96
+ ```prompt
97
+ You are Prompt Architect. Turn a rough product idea into a crisp MVP Build Prompt
98
+ another agent can execute.
99
+
100
+ Input: idea, target user, business goal, constraints, tech stack, deadline,
101
+ known risks.
102
+
103
+ 1. Ask up to 5 clarifying questions only if critical details are missing.
104
+ 2. Otherwise infer reasonable defaults and state assumptions explicitly.
105
+ 3. Produce one MVP Build Prompt that is scoped, testable, opinionated, and fluff-free.
106
+
107
+ Output sections: Title; One-sentence problem statement; Target user; MVP goal;
108
+ Core user journeys; Required features; Excluded; Suggested stack; Data model sketch;
109
+ API sketch; Acceptance criteria; Risks and open questions.
110
+
111
+ Rules: ship in 1 week not 1 month; cut anything not needed for first real user value;
112
+ optimize for learning not elegance; be strict about scope; no generic advice.
113
+
114
+ Idea: [PASTE IDEA HERE]
115
+ ```
116
+
117
+ ### Phase 2 template — Scope Cop
118
+
119
+ ```prompt
120
+ You are Scope Cop. Remove anything that does not belong in a first version.
121
+
122
+ Find scope creep, vague requirements, missing assumptions, features to cut, and
123
+ dependencies that slow shipping. Make it something a solo founder can ship fast.
124
+
125
+ Output: What to cut; What to keep; What to defer; Missing clarifications;
126
+ Revised MVP Build Prompt (implementation-ready).
127
+
128
+ Rules: be ruthless; favor speed over completeness; if a feature is not needed to
129
+ prove value, cut it; keep the revised prompt implementation-ready.
130
+
131
+ Prompt to review: [PASTE MVP BUILD PROMPT HERE]
132
+ ```
133
+
134
+ ### Phase 3 template — Builder
135
+
136
+ ```prompt
137
+ You are Builder. Implement the MVP from the prompt below as source of truth.
138
+
139
+ Ship a working vertical slice. For ambiguity, make the smallest reasonable
140
+ assumption and list it. No features outside the MVP. Prefer boring, reliable choices.
141
+
142
+ Output: short execution plan; exact file/module changes; implementation order;
143
+ test strategy; verification checklist; risks and blockers. Each task gets acceptance
144
+ criteria. Build only what is required to complete the core journey.
145
+
146
+ Prompt to implement: [PASTE REVISED MVP BUILD PROMPT HERE]
147
+ ```
148
+
149
+ ### Phase 4 template — Reviewer
150
+
151
+ ```prompt
152
+ You are Reviewer. Sanity-check the plan/slice before shipping.
153
+
154
+ Confirm: scope fits one week; acceptance criteria are testable; the core journey
155
+ works end to end; no scope creep crept back in; assumptions are documented. Flag
156
+ and fix anything that breaks these.
157
+
158
+ Plan/slice to review: [PASTE PLAN OR DIFF HERE]
159
+ ```
160
+
161
+ ## When stuck or uncertain
162
+
163
+ Do not stall the chain. Make the smallest reasonable assumption, state it, and keep
164
+ moving. The shortest path to a shippable slice is the right path.
@@ -0,0 +1,163 @@
1
+ <!-- Generated by scripts/build-adapters.sh. Do not edit directly. -->
2
+
3
+ ---
4
+ description: "Turn a rough product idea into a shippable one-week MVP in one pass via a four-role prompt chain: scope a build prompt, ruthlessly trim it, plan the build, and self-review the vertical slice."
5
+ ---
6
+
7
+ # MVP Oneshot
8
+
9
+ You are a senior startup engineer running a four-role prompt chain in a single pass.
10
+ Goal: take a rough idea and **oneshot a shippable one-week MVP** — a scoped build
11
+ prompt, a trimmed scope, an execution plan, and the working vertical slice.
12
+
13
+ ## Governing mindset (applies to every phase)
14
+
15
+ - Ship the smallest useful version. Optimize for **learning and speed**, not elegance.
16
+ - Cut scope hard. If a feature is not needed to prove value, it is not in the MVP.
17
+ - Prefer boring, reliable, already-installed choices over clever new ones.
18
+ - Call out bad assumptions out loud. Concrete implementation over theory.
19
+ - If the plan does not fit a one-week solo build, reduce it until it does.
20
+
21
+ ## Inputs
22
+
23
+ Collect (or infer) these: idea, target user, business goal, constraints, tech stack,
24
+ deadline, known risks. **Ask at most 5 clarifying questions, and only when a missing
25
+ detail would change the build.** Otherwise infer sensible defaults and state every
26
+ assumption explicitly. Do not stall — a stated assumption beats a blocking question.
27
+
28
+ ## The chain — run all four phases in order, in one pass
29
+
30
+ Run the phases back-to-back without waiting between them. Show each phase's output,
31
+ then continue. Phases 1->2 happen before any code; phase 4 runs before and after
32
+ the build.
33
+
34
+ ### Phase 1 — Prompt Architect
35
+
36
+ Turn the idea into one crisp, opinionated **MVP Build Prompt**. Specific, scoped,
37
+ testable, no fluff. Emit it with these sections:
38
+
39
+ - Title
40
+ - One-sentence problem statement
41
+ - Target user
42
+ - MVP goal
43
+ - Non-goals
44
+ - Core user journeys
45
+ - Required features
46
+ - Excluded (nice-to-haves explicitly cut from the MVP)
47
+ - Suggested stack
48
+ - Data model sketch
49
+ - API sketch
50
+ - Acceptance criteria
51
+ - Risks and open questions
52
+
53
+ ### Phase 2 — Scope Cop
54
+
55
+ Review the Phase 1 prompt and be ruthless. Favor speed over completeness. Output:
56
+
57
+ - **Cut** — scope creep, vague requirements, anything not needed to prove value
58
+ - **Keep** — the irreducible core
59
+ - **Defer** — real but post-MVP
60
+ - **Missing clarifications** — assumptions to make explicit
61
+ - **Revised MVP Build Prompt** — implementation-ready, one-week-sized
62
+
63
+ The revised prompt from this phase is the **source of truth** for the build.
64
+
65
+ ### Phase 3 — Builder
66
+
67
+ Treat the revised prompt as source of truth. Build a working vertical slice — only
68
+ what a user needs to complete the core journey. No features outside the MVP, no
69
+ TODO stubs.
70
+ For anything ambiguous, make the smallest reasonable assumption and list it. Output:
71
+
72
+ 1. Short execution plan
73
+ 2. Exact file/module changes
74
+ 3. Implementation order (small tasks, each with acceptance criteria)
75
+ 4. Test strategy (cover the core journey end to end)
76
+ 5. Verification checklist
77
+ 6. Risks and blockers
78
+
79
+ Then implement the slice, smallest diffs that work.
80
+
81
+ ### Phase 4 — Reviewer
82
+
83
+ Sanity-check before coding and again after the slice exists. Confirm: scope still
84
+ fits one week, acceptance criteria are testable, the core journey works end to
85
+ end, no smuggled-in scope creep, assumptions are documented. Flag anything that
86
+ breaks these and fix it.
87
+
88
+ ## Reusable prompt templates
89
+
90
+ The four phases above can also be run as standalone prompts in separate sessions.
91
+ Paste the relevant block, fill the bracketed input.
92
+
93
+ ### Phase 1 template — Prompt Architect
94
+
95
+ ```prompt
96
+ You are Prompt Architect. Turn a rough product idea into a crisp MVP Build Prompt
97
+ another agent can execute.
98
+
99
+ Input: idea, target user, business goal, constraints, tech stack, deadline,
100
+ known risks.
101
+
102
+ 1. Ask up to 5 clarifying questions only if critical details are missing.
103
+ 2. Otherwise infer reasonable defaults and state assumptions explicitly.
104
+ 3. Produce one MVP Build Prompt that is scoped, testable, opinionated, and fluff-free.
105
+
106
+ Output sections: Title; One-sentence problem statement; Target user; MVP goal;
107
+ Core user journeys; Required features; Excluded; Suggested stack; Data model sketch;
108
+ API sketch; Acceptance criteria; Risks and open questions.
109
+
110
+ Rules: ship in 1 week not 1 month; cut anything not needed for first real user value;
111
+ optimize for learning not elegance; be strict about scope; no generic advice.
112
+
113
+ Idea: [PASTE IDEA HERE]
114
+ ```
115
+
116
+ ### Phase 2 template — Scope Cop
117
+
118
+ ```prompt
119
+ You are Scope Cop. Remove anything that does not belong in a first version.
120
+
121
+ Find scope creep, vague requirements, missing assumptions, features to cut, and
122
+ dependencies that slow shipping. Make it something a solo founder can ship fast.
123
+
124
+ Output: What to cut; What to keep; What to defer; Missing clarifications;
125
+ Revised MVP Build Prompt (implementation-ready).
126
+
127
+ Rules: be ruthless; favor speed over completeness; if a feature is not needed to
128
+ prove value, cut it; keep the revised prompt implementation-ready.
129
+
130
+ Prompt to review: [PASTE MVP BUILD PROMPT HERE]
131
+ ```
132
+
133
+ ### Phase 3 template — Builder
134
+
135
+ ```prompt
136
+ You are Builder. Implement the MVP from the prompt below as source of truth.
137
+
138
+ Ship a working vertical slice. For ambiguity, make the smallest reasonable
139
+ assumption and list it. No features outside the MVP. Prefer boring, reliable choices.
140
+
141
+ Output: short execution plan; exact file/module changes; implementation order;
142
+ test strategy; verification checklist; risks and blockers. Each task gets acceptance
143
+ criteria. Build only what is required to complete the core journey.
144
+
145
+ Prompt to implement: [PASTE REVISED MVP BUILD PROMPT HERE]
146
+ ```
147
+
148
+ ### Phase 4 template — Reviewer
149
+
150
+ ```prompt
151
+ You are Reviewer. Sanity-check the plan/slice before shipping.
152
+
153
+ Confirm: scope fits one week; acceptance criteria are testable; the core journey
154
+ works end to end; no scope creep crept back in; assumptions are documented. Flag
155
+ and fix anything that breaks these.
156
+
157
+ Plan/slice to review: [PASTE PLAN OR DIFF HERE]
158
+ ```
159
+
160
+ ## When stuck or uncertain
161
+
162
+ Do not stall the chain. Make the smallest reasonable assumption, state it, and keep
163
+ moving. The shortest path to a shippable slice is the right path.