03-planning-and-research

Name: 03-planning-and-research
Author: heymegabyte

18stars4forks

How to install 03-planning-and-research?

heymegabyte/claude-skills/03-planning-and-research

$npx -y skills add heymegabyte/claude-skills --skill 03-planning-and-research

Installs into the current project.

Summary

Deep web research, competitor scanning, technology evaluation, and implementation planning. Decomposes work into vertical slices, identifies parallel workstreams, tracks assumptions with confidence levels, and designs the critical path for minimum wall-clock time.

Files · 1

View on GitHub

SKILL.md

1	# 03 — Planning and Research
2
3	Decompose work into parallel vertical slices using deep web research, competitor scanning, and technology evaluation.
4
5	## Deep research protocol
6
7	Before any non-trivial implementation:
8
9	1. Web search top 50 results for primary keyphrase
10	2. Read top 10 in depth — extract directives + counter-arguments
11	3. Cross-ref against existing rule mesh
12	4. Confidence-track every claim 0-1
13
14	Use `web_search_20260209` + `web_fetch_20260209` (free when paired with `code_execution_20260120`).
15
16	## Competitor scanning
17
18	For every website build, run `rules/competitor-research.md` Phase -1 BEFORE Phase 0:
19
20	- Identify top 5-10 audience-comparable sites
21	- Score each on 100-pt rubric (10 dims × 10pts)
22	- Set ≥15% beat-floor for Phase 6 loop termination
23
24	## Technology evaluation
25
26	For every new dep / framework / service consideration:
27
28	1. Already have equivalent? → use it
29	2. Feature truly needed? → if speculative, defer
30	3. License? → OSS free only (per `rules/brian-preferences.md` priority order)
31	4. CF compat? → adapter pattern per `rules/cloudflare-hostable-supervisor.md`
32	5. Bundle/perf impact? → measure before commit
33	6. Lighter existing solution? → prefer
34	7. Decide: install now · defer · adapter-only · reject + document why
35
36	## Implementation planning
37
38	### Decomposition
39
40	- Identify atomic units (file × layer)
41	- Mark dependencies between units
42	- Distinguish independent vs serial chains
43
44	### Vertical slicing
45
46	Per `06-build-and-slice-loop`:
47
48	- Each slice ships through every layer (UI → API → DB → tests → deploy)
49	- Homepage FIRST (no exceptions)
50	- Slice = one feature, not one layer
51
52	### Parallelism plan
53
54	Per `rules/parallel-subagent-economy.md`:
55
56	- ≥5-min wall-clock saving + independent → fan out
57	- Sweet spot 3-4 specialists, ceiling 6
58	- Batch beyond 6 in waves
59	- Sonnet default; Opus for security/architect/visual-qa overrides
60
61	### Critical path
62
63	- Identify longest dependency chain
64	- Front-load research that gates implementation
65	- Parallelize everything off the critical path
66	- Wall-clock = max(critical_path, max parallel branch)
67
68	## Assumption tracking
69
70	Every assumption logged in `_assumptions.md` with:
71
72	- Claim
73	- Confidence 0-1
74	- Evidence sources
75	- Decision impact if wrong
76	- Verification trigger
77
78	Confidence <0.7 → research more. Per `rules/auto-meta-work.md`.
79
80	## Three-Layer Knowledge
81
82	Per `~/.claude/CLAUDE.md` § Thinking:
83
84	- L1 = proven (existing rules + patterns)
85	- L2 = trending (blog posts, recent papers)
86	- L3 = first principles (build from scratch)
87
88	Prefer L3. Best outcome of research is NOT finding a solution to copy — it's understanding the problem deeply enough to design a better one.
89
90	## Self-Argue (before major decisions)
91
92	Generate strongest counterargument. If you can't defeat it, decision is wrong.
93
94	## Boil-the-lake
95
96	Marginal cost of completeness is near-zero. When complete costs minutes more than shortcut, do complete. Boil lakes, flag oceans.
97
98	## Output artifacts
99
100	- `_research.json` — raw findings, source URLs, confidence
101	- `_assumptions.md` — tracked claims
102	- `PLAN.md` — implementation roadmap w/ parallelism plan + critical path
103	- `_decisions.md` — architectural decisions w/ rationale + alternatives
104	- `_brief_summary.txt` — 100-word digest for downstream agents

People who liked this also viewed

6 picks

Type

03-planning-and-research

18stars4forks

View on GitHub

How to install 03-planning-and-research?

heymegabyte/claude-skills/03-planning-and-research

$npx -y skills add heymegabyte/claude-skills --skill 03-planning-and-research

Installs into the current project.

Summary

Files · 1

View on GitHub

SKILL.md

1	# 03 — Planning and Research
2
3	Decompose work into parallel vertical slices using deep web research, competitor scanning, and technology evaluation.
4
5	## Deep research protocol
6
7	Before any non-trivial implementation:
8
9	1. Web search top 50 results for primary keyphrase
10	2. Read top 10 in depth — extract directives + counter-arguments
11	3. Cross-ref against existing rule mesh
12	4. Confidence-track every claim 0-1
13
14	Use `web_search_20260209` + `web_fetch_20260209` (free when paired with `code_execution_20260120`).
15
16	## Competitor scanning
17
18	For every website build, run `rules/competitor-research.md` Phase -1 BEFORE Phase 0:
19
20	- Identify top 5-10 audience-comparable sites
21	- Score each on 100-pt rubric (10 dims × 10pts)
22	- Set ≥15% beat-floor for Phase 6 loop termination
23
24	## Technology evaluation
25
26	For every new dep / framework / service consideration:
27
28	1. Already have equivalent? → use it
29	2. Feature truly needed? → if speculative, defer
30	3. License? → OSS free only (per `rules/brian-preferences.md` priority order)
31	4. CF compat? → adapter pattern per `rules/cloudflare-hostable-supervisor.md`
32	5. Bundle/perf impact? → measure before commit
33	6. Lighter existing solution? → prefer
34	7. Decide: install now · defer · adapter-only · reject + document why
35
36	## Implementation planning
37
38	### Decomposition
39
40	- Identify atomic units (file × layer)
41	- Mark dependencies between units
42	- Distinguish independent vs serial chains
43
44	### Vertical slicing
45
46	Per `06-build-and-slice-loop`:
47
48	- Each slice ships through every layer (UI → API → DB → tests → deploy)
49	- Homepage FIRST (no exceptions)
50	- Slice = one feature, not one layer
51
52	### Parallelism plan
53
54	Per `rules/parallel-subagent-economy.md`:
55
56	- ≥5-min wall-clock saving + independent → fan out
57	- Sweet spot 3-4 specialists, ceiling 6
58	- Batch beyond 6 in waves
59	- Sonnet default; Opus for security/architect/visual-qa overrides
60
61	### Critical path
62
63	- Identify longest dependency chain
64	- Front-load research that gates implementation
65	- Parallelize everything off the critical path
66	- Wall-clock = max(critical_path, max parallel branch)
67
68	## Assumption tracking
69
70	Every assumption logged in `_assumptions.md` with:
71
72	- Claim
73	- Confidence 0-1
74	- Evidence sources
75	- Decision impact if wrong
76	- Verification trigger
77
78	Confidence <0.7 → research more. Per `rules/auto-meta-work.md`.
79
80	## Three-Layer Knowledge
81
82	Per `~/.claude/CLAUDE.md` § Thinking:
83
84	- L1 = proven (existing rules + patterns)
85	- L2 = trending (blog posts, recent papers)
86	- L3 = first principles (build from scratch)
87
88	Prefer L3. Best outcome of research is NOT finding a solution to copy — it's understanding the problem deeply enough to design a better one.
89
90	## Self-Argue (before major decisions)
91
92	Generate strongest counterargument. If you can't defeat it, decision is wrong.
93
94	## Boil-the-lake
95
96	Marginal cost of completeness is near-zero. When complete costs minutes more than shortcut, do complete. Boil lakes, flag oceans.
97
98	## Output artifacts
99
100	- `_research.json` — raw findings, source URLs, confidence
101	- `_assumptions.md` — tracked claims
102	- `PLAN.md` — implementation roadmap w/ parallelism plan + critical path
103	- `_decisions.md` — architectural decisions w/ rationale + alternatives
104	- `_brief_summary.txt` — 100-word digest for downstream agents

People who liked this also viewed

6 picks

Type