market-researcher

Name: market-researcher
Author: tronghieu

37stars20forks

How to install market-researcher?

tronghieu/agent-skills/market-researcher

$npx -y skills add tronghieu/agent-skills --skill market-researcher

Installs into the current project.

Summary

Conduct rigorous, citation-backed desk market research: quick go/no-go market scans, market sizing (TAM/SAM/SOM), competitor analysis, customer demand signals, and trend/macro (PESTEL) analysis. Use this skill whenever the user wants to research, size, validate, or enter a market

Files · 1

View on GitHub

SKILL.md

1	# Market Researcher
2
3	A desk-research engine for market questions. It takes a raw question like
4	"should we build X?" or "how big is the market for Y in Brazil?" and returns a
5	research report in which every factual claim traces to a registered source
6	— the discipline the typical AI market analysis lacks, and the reason this
7	skill exists. A confident fabricated market size looks exactly like a real one;
8	the citation machinery below is what makes the difference detectable.
9
10	The skill runs desk research only: web sources, public data, documents the
11	user provides. It designs around that limit honestly — what it learns about
12	customers are demand signals and hypotheses, never validated insights (see
13	non-negotiable 5).
14
15	## Non-negotiables
16
17	These five rules outrank everything else in this skill:
18
19	1. Every claim is traceable. Every number and every factual assertion in a
20	deliverable either carries a source tag `[S#]` that resolves to a row in
21	`sources.md`, or is an explicit derived estimate showing its formula and
22	cited inputs, or is labelled `(assumption A#)`. Nothing rides on authority
23	of tone. If you catch yourself writing "the market is growing rapidly"
24	without a tag, stop and either find the source or cut the sentence.
25	2. Current facts come from the web, not from memory. Market sizes, prices,
26	competitor products, funding, regulations — research them fresh. Training
27	data is stale and, worse, statistically plausible. If web access is
28	unavailable, say so and stop; do not degrade into an uncited essay.
29	3. Triangulate; never silently average. Size markets with more than one
30	method when the decision warrants it. When two credible sources or methods
31	disagree by more than ~2x, report the conflict and diagnose it (different
32	market definitions, years, currencies, revenue-vs-GMV) rather than blending
33	it away.
34	4. A verification pass runs before delivery. A skeptic re-checks key
35	claims against their sources, re-does estimate arithmetic, and flags stale
36	data. Findings it cannot resolve ship in the report as caveats, not
37	silently dropped.
38	5. Desk research yields signals, not validated customer insight. Personas
39	and needs produced here are hypotheses to be tested by real primary
40	research (interviews, tests) that the user or a calling skill owns. Label
41	them as such and always end customer sections with "open questions for
42	primary research".
43
44	## Prerequisite
45
46	Web search and web fetch. Check availability before promising anything; if
47	missing, tell the user what you need and stop rather than answering from
48	training data (non-negotiable 2).
49
50	### Tool economy
51
52	Desk research is reading, not browsing. Use the cheapest tool that gets the
53	text, in this order:
54
55	1. Built-in web search (WebSearch or equivalent) to find sources.
56	2. Built-in fetch (WebFetch or equivalent) to read them — it returns the
57	page text, which is all a citation needs.
58	3. Browser automation (agent-browser, Chrome/Playwright tools) only as a
59	last resort for a specific page that fetch genuinely cannot read
60	(JS-only rendering, interaction-gated content) and that no alternative
61	source covers. Never screenshot pages to "read" them.
62
63	This ordering is a cost rule, not a style preference: a browser session burns
64	tokens on screenshots and DOM snapshots at a rate that can exceed the entire
65	rest of the lane, and it produces nothing a citation needs that fetched text
66	doesn't. If another skill's guidance says to prefer browser tools, it does
67	not apply inside this skill's research lanes. Pass this rule to every lane
68	subagent you spawn.
69
70	One harness quirk: some environments guard the Write tool against
71	report-shaped filenames (`report.md`, `summary.md`, `findings-*.md`). If a
72	required output file is blocked, write it via a shell heredoc instead — the
73	filenames are part of this skill's contract and must not be renamed to
74	appease the guard.
75
76	## Two modes
77
78	\| \| Quick Scan (default) \| Deep Dive (opt-in) \|
79	\|---\|---\|---\|
80	\| Question \| "Is this worth a closer look?" \| "What exactly are we wal

People who liked this also viewed

6 picks

Type

market-researcher

37stars20forks

View on GitHub

How to install market-researcher?

tronghieu/agent-skills/market-researcher

$npx -y skills add tronghieu/agent-skills --skill market-researcher

Installs into the current project.

Summary

Files · 1

View on GitHub

SKILL.md

1	# Market Researcher
2
3	A desk-research engine for market questions. It takes a raw question like
4	"should we build X?" or "how big is the market for Y in Brazil?" and returns a
5	research report in which every factual claim traces to a registered source
6	— the discipline the typical AI market analysis lacks, and the reason this
7	skill exists. A confident fabricated market size looks exactly like a real one;
8	the citation machinery below is what makes the difference detectable.
9
10	The skill runs desk research only: web sources, public data, documents the
11	user provides. It designs around that limit honestly — what it learns about
12	customers are demand signals and hypotheses, never validated insights (see
13	non-negotiable 5).
14
15	## Non-negotiables
16
17	These five rules outrank everything else in this skill:
18
19	1. Every claim is traceable. Every number and every factual assertion in a
20	deliverable either carries a source tag `[S#]` that resolves to a row in
21	`sources.md`, or is an explicit derived estimate showing its formula and
22	cited inputs, or is labelled `(assumption A#)`. Nothing rides on authority
23	of tone. If you catch yourself writing "the market is growing rapidly"
24	without a tag, stop and either find the source or cut the sentence.
25	2. Current facts come from the web, not from memory. Market sizes, prices,
26	competitor products, funding, regulations — research them fresh. Training
27	data is stale and, worse, statistically plausible. If web access is
28	unavailable, say so and stop; do not degrade into an uncited essay.
29	3. Triangulate; never silently average. Size markets with more than one
30	method when the decision warrants it. When two credible sources or methods
31	disagree by more than ~2x, report the conflict and diagnose it (different
32	market definitions, years, currencies, revenue-vs-GMV) rather than blending
33	it away.
34	4. A verification pass runs before delivery. A skeptic re-checks key
35	claims against their sources, re-does estimate arithmetic, and flags stale
36	data. Findings it cannot resolve ship in the report as caveats, not
37	silently dropped.
38	5. Desk research yields signals, not validated customer insight. Personas
39	and needs produced here are hypotheses to be tested by real primary
40	research (interviews, tests) that the user or a calling skill owns. Label
41	them as such and always end customer sections with "open questions for
42	primary research".
43
44	## Prerequisite
45
46	Web search and web fetch. Check availability before promising anything; if
47	missing, tell the user what you need and stop rather than answering from
48	training data (non-negotiable 2).
49
50	### Tool economy
51
52	Desk research is reading, not browsing. Use the cheapest tool that gets the
53	text, in this order:
54
55	1. Built-in web search (WebSearch or equivalent) to find sources.
56	2. Built-in fetch (WebFetch or equivalent) to read them — it returns the
57	page text, which is all a citation needs.
58	3. Browser automation (agent-browser, Chrome/Playwright tools) only as a
59	last resort for a specific page that fetch genuinely cannot read
60	(JS-only rendering, interaction-gated content) and that no alternative
61	source covers. Never screenshot pages to "read" them.
62
63	This ordering is a cost rule, not a style preference: a browser session burns
64	tokens on screenshots and DOM snapshots at a rate that can exceed the entire
65	rest of the lane, and it produces nothing a citation needs that fetched text
66	doesn't. If another skill's guidance says to prefer browser tools, it does
67	not apply inside this skill's research lanes. Pass this rule to every lane
68	subagent you spawn.
69
70	One harness quirk: some environments guard the Write tool against
71	report-shaped filenames (`report.md`, `summary.md`, `findings-*.md`). If a
72	required output file is blocked, write it via a shell heredoc instead — the
73	filenames are part of this skill's contract and must not be renamed to
74	appease the guard.
75
76	## Two modes
77
78	\| \| Quick Scan (default) \| Deep Dive (opt-in) \|
79	\|---\|---\|---\|
80	\| Question \| "Is this worth a closer look?" \| "What exactly are we wal

People who liked this also viewed

6 picks

Type