| name | market-intelligence-assessment |
| description | Mandatory market intelligence assessment that runs before any spec work begins. Researches competitive landscape, validates customer pain, assesses market size, identifies differentiation opportunities, and flags risk signals. Produces a GO / CAUTION / STOP recommendation for human review.
|
| agents | {"primary":"business-analyst","supporting":["research-agent"]} |
| inputs | ["Human's initial input (idea, brief, PRD, or full spec)",".factory/planning/artifact-inventory.md (from artifact detection)"] |
| outputs | [".factory/planning/market-intel.md"] |
| gate | Human reviews market intel before proceeding |
Market Intelligence Assessment
Purpose
Every product idea, feature request, or spec entering the Dark Factory pipeline must
pass through a market intelligence assessment before any spec work begins. This skill
ensures the factory never invests engineering effort into products or features that:
- Already exist with strong incumbents and no differentiation angle
- Solve problems customers don't actually have (or don't pay to solve)
- Target markets too small to justify the investment
- Carry unidentified regulatory, competitive, or timing risks
The assessment produces a GO / CAUTION / STOP recommendation. The human always makes
the final call -- the factory provides the evidence.
Step 1: Extract Market Context
Agent: business-analyst
The business-analyst reads whatever the human provided (L0 idea, L1 brief, L2 PRD,
L3 architecture, or L4 full spec) and extracts a structured research brief:
## Research Brief
- **Product/Feature:** [one-sentence description]
- **Target Audience:** [who this is for, as specific as available]
- **Problem Statement:** [what pain this solves]
- **Proposed Solution:** [how it solves the pain, high-level]
- **Known Competitors:** [any competitors the human mentioned]
- **Known Constraints:** [regulatory, technical, market constraints mentioned]
- **Input Level:** [L0/L1/L2/L3/L4]
The research brief is passed to the research-agent for Step 2.
If the input is sparse (L0 idea with minimal detail), the business-analyst asks the
human up to 3 clarifying questions before proceeding. Do not block on answers -- use
reasonable assumptions and flag them as assumptions in the output.
Step 2: Market Landscape Research
Agent: research-agent (canonical MCP caller; this skill spawns it)
The research-agent executes five parallel research tracks, leading with deep research (perplexity_research, backed by sonar-deep-research, reasoning_effort: high for this landscape-level work):
2a. Competitive Landscape
- Direct competitors (same problem, same audience)
- Adjacent competitors (similar problem, different audience or approach)
- Emerging competitors (startups, open-source projects, research papers)
- For each competitor: name, approach, funding/traction, strengths, weaknesses
2b. Market Size
- TAM/SAM/SOM estimates with sources
- Growth rate and trajectory
- Pricing benchmarks from existing solutions
- Market maturity signal (nascent / growing / mature / declining)
2c. Customer Pain Validation
- Evidence that the target audience experiences this pain (forums, reviews, surveys)
- How they currently solve it (workarounds, manual processes, existing tools)
- Willingness to pay signals (existing spending on workarounds, budget categories)
- Pain severity: inconvenience vs. blocker vs. revenue-impacting
2d. Differentiation Opportunities
- Gaps in existing solutions (features, UX, pricing, integration, performance)
- Underserved segments within the target audience
- Technology shifts enabling new approaches (AI, new protocols, regulatory changes)
- Timing advantages (why now, not two years ago?)
2e. Risk Signals
- Regulatory risks (pending legislation, compliance requirements)
- Platform risks (dependency on APIs, marketplaces, ecosystems that could change)
- Competitive moat risks (can incumbents easily replicate the differentiation?)
- Market timing risks (too early, too late, window closing)
- Technology risks (depends on immature tech, unproven at scale)
Step 3: Synthesis and Recommendation
Agent: business-analyst
The business-analyst synthesizes research findings into .factory/planning/market-intel.md
using the following template:
---
recommendation: GO | CAUTION | STOP
confidence: high | medium | low
input_level: L0 | L1 | L2 | L3 | L4
assessed_at: [ISO timestamp]
assessor: business-analyst + research-agent
---
# Market Intelligence Assessment
## Executive Summary
[2-3 sentences: what we found, what it means, what we recommend]
## 1. Competitive Landscape
### Direct Competitors
| Competitor | Approach | Traction | Strengths | Weaknesses |
|-----------|----------|----------|-----------|------------|
| ... | ... | ... | ... | ... |
### Adjacent Competitors
[same format]
### Emerging Threats
[same format]
### Competitive Density Score
[LOW / MEDIUM / HIGH / SATURATED]
## 2. Market Size & Dynamics
- **TAM:** [estimate + source]
- **SAM:** [estimate + source]
- **SOM:** [estimate + source]
- **Growth Rate:** [X% CAGR, source]
- **Market Maturity:** [nascent / growing / mature / declining]
- **Pricing Benchmarks:** [range from existing solutions]
## 3. Customer Pain Validation
- **Pain Confirmed:** [YES / PARTIAL / NO]
- **Evidence:** [specific sources -- forums, reviews, surveys, interviews]
- **Current Workarounds:** [how people solve this today]
- **Willingness to Pay:** [signals for/against]
- **Pain Severity:** [inconvenience / blocker / revenue-impacting]
## 4. Differentiation Opportunities
- [opportunity 1: description + evidence]
- [opportunity 2: description + evidence]
- [opportunity 3: description + evidence]
## 5. Risk Signals
| Risk | Severity | Likelihood | Mitigation |
|------|----------|-----------|------------|
| ... | HIGH/MED/LOW | HIGH/MED/LOW | ... |
## 6. Implications for Spec Work
- **If GO:** [what the spec should emphasize, what to prioritize]
- **If CAUTION:** [what needs more research, what assumptions to test]
- **If STOP:** [why, what would need to change for GO]
### Assumptions Made
- [assumption 1 -- flagged for human validation]
- [assumption 2 -- flagged for human validation]
Recommendation Criteria
- GO: Pain confirmed, market viable, clear differentiation exists, risks manageable
- CAUTION: Pain partially confirmed OR differentiation unclear OR significant risks
identified. Proceed with awareness -- spec should address the flagged concerns.
- STOP: Pain not validated, market too small, saturated with no differentiation,
or showstopper risks identified. Human should reconsider before investing spec effort.
Confidence level reflects research depth:
- high: Multiple corroborating sources, recent data, clear signals
- medium: Some sources, some assumptions, mixed signals
- low: Limited data available, significant assumptions made
Step 4: Human Review Gate
The market intelligence assessment is presented to the human for review.
- GO: Human approves, pipeline continues to spec work
- CAUTION: Human reviews flagged concerns, provides direction (proceed, pivot, research more)
- STOP: Human reviews reasoning, decides to abort, pivot, or override with justification
The human's decision is recorded in STATE.md:
market_intel:
recommendation: GO | CAUTION | STOP
human_decision: approved | overridden | pivoted | aborted
notes: "[human's reasoning if overriding]"
If the human overrides a STOP recommendation, the override and reasoning are carried
forward as context for all downstream agents (they should know the market risk was
accepted deliberately).
Depth Configuration
The assessment depth adapts based on the input level (L0-L4) to avoid redundant
research on projects that already have market validation:
L0: Full Landscape (idea, no artifacts)
- Sections: All 5 (Competitive Landscape, Market Size, Customer Pain Validation,
Differentiation Opportunities, Risk Signals)
- Depth: Maximum -- this is the first time anyone has researched this idea
- Expected Duration: 15-30 minutes
L1: Validation Focus (brief exists)
- Sections: 1 (Competitive Landscape), 3 (Customer Pain Validation), 5 (Risk Signals)
- Depth: Validate claims made in the brief against external evidence
- Expected Duration: 10-15 minutes
L2: Competitive Deep-Dive (PRD exists)
- Sections: 1 (Competitive Landscape), 4 (Differentiation Opportunities), 5 (Risk Signals)
- Depth: PRD implies pain is validated; focus on competitive positioning and risks
- Expected Duration: 10-15 minutes
L3: Risk Check (architecture exists)
- Sections: 5 (Risk Signals) + summary of 1 (Competitive Landscape)
- Depth: Architecture implies deep investment; check for late-stage risks only
- Expected Duration: 5-10 minutes
L4: Freshness Check (full spec exists)
- Sections: Headline check of 1 (Competitive Landscape) + new risks in 5 (Risk Signals)
- Depth: Spec is complete; only check if the landscape changed since spec was written
- Auto-GO: If no material changes detected, auto-recommend GO without human gate
- Expected Duration: 2-5 minutes
Quality Gate
Failure Modes
- If research sources are sparse (few results, low confidence): produce CAUTION recommendation with explicit gaps listed
- If Perplexity MCP is unavailable: do not silently skip — per the research-agent's mandatory-MCP gate, escalate the toolchain failure with the verbatim error, then fall back to training data with an "UNVERIFIED -- no live research" disclaimer and default to CAUTION
- If the input is too vague for meaningful research (L0 with one sentence): ask up to 3 clarifying questions, then proceed with assumptions flagged
