| name | edge-profiler |
| description | Extracts and structures a user's skills, assets, access, constraints, credibility, and usable versus restricted edge into a reusable founder profile. Use when planning businesses, ranking opportunities, validating founder-market fit, or whenever later skills need a clean profile instead of raw conversation. |
| compatibility | Works best with read/search tools and benefits from conversational follow-up when user context is thin. |
| metadata | {"author":"opencode","version":"0.1"} |
Edge Profiler
Overview
This skill turns messy user background into a reusable founder-edge profile. The
goal is not to flatter the user. The goal is to separate real edge from weak,
restricted, or unusable edge so later skills can reason from something solid.
Treat edge as four different things, not one bucket:
usable_edge
restricted_or_unusable_edge
fake_or_overstated_edge
missing_edge
Load references/profile-dimensions.md when filling the profile in detail. Load
references/intake-questions.md when the user context is thin or vague.
When to use
- When a user wants business ideas, monetization paths, or opportunity ranking
- When a later skill needs a clean profile instead of raw conversation history
- When the user gives a long background and the important advantages need to be
extracted cleanly
- When founder-market fit, distribution quality, or constraint realism are
central to the task
Do NOT use when:
- The user already has a clean structured profile and only needs execution on a
chosen idea
- The task is purely implementation and the user's background is irrelevant
Workflow
1. Extract evidence from the conversation
Pull concrete evidence from what the user actually said. Do not invent stronger
credentials, access, or buyer trust than the prompt supports.
Capture at least:
- Skills
- Domain experience
- Assets
- Distribution
- Network access
- Constraints
- Credibility signals
- Unfair advantages
2. Distinguish usable edge from restricted edge
Separate the profile into:
usable_edge
restricted_or_unusable_edge
Restricted or unusable edge includes things like:
- Employer-confidential information
- Customer access the user cannot legally monetize
- Regulated authority the user does not control
- Weak signals dressed up as strong edge
If the user's only apparent moat is restricted, say so plainly.
3. Identify fake or overstated edge
Some signals sound impressive but do not create commercial leverage by
themselves.
Common examples:
- follower count without buyer quality
- employer-mediated access that is not independently owned
- prestige brands or smart peers without route to customers
- broad market familiarity without trust, access, or distribution
Do not inflate these into moat. Put them under fake_or_overstated_edge or
weak_spots.
4. Fill gaps without flattering
If the profile is incomplete, ask up to 5 short follow-up questions.
If niche, buyer access, or distribution quality is weak or unclear, spend most
of those questions on wedge discovery before concluding what the user's real
usable edge is. Do not default to assumptions when the missing information is
the route to customers.
Only proceed with assumptions when follow-up is genuinely impossible or the user
explicitly declines more questions. In that case, state clearly that the profile
is provisional and that any opportunity recommendation should be treated as
lower confidence.
Do not confuse:
- follower count with buyer quality
- industry familiarity with permission to act
- technical skill with distribution
- interest with edge
- prestige with buyer access
- introductions with durable distribution
For every claimed edge, pressure-test it with these questions:
- Is it owned?
- Is it legal to use?
- Is it buyer-relevant?
- Is it repeatable?
- Is it independent of the user's employer?
Mandatory deeper intake trigger:
If the user has strong skills but weak or unclear niche, buyer access, or
distribution, ask targeted follow-up in these categories before finalizing the
profile:
- reachable buyer segment
- repeated painful workflow
- prior trust, referrals, or pull signals
- owned distribution or warm-path quality
- clean-room-safe relationships and assets
- willingness to start with a service or concierge wedge
Keep this to 3-5 total questions, but make them decision-critical. Prefer
questions that collapse uncertainty about who the first buyer is, why they would
care, and how the user can reach them legally and independently.
5. Produce the reusable profile
Return both:
- a human-readable profile summary
- a JSON block using
assets/profile-template.json
Use the template field names exactly. This skill is meant to feed later skills,
so schema consistency matters.
Default human-readable sections:
Profile Summary
Usable Edge
Restricted Edge
Fake or Overstated Edge
Weak Spots
Follow-Up Questions
Structured Profile
Substance requirements:
- Identify at least one real advantage or say the edge is weak
- Identify restricted edge when it exists
- Be explicit about buyer/distribution quality, not just audience size
- Name the top missing information if it materially affects later decisions
- If something sounds impressive but is not monetizable, call that out directly
- Keep restricted materials out of
user_profile.assets; put them only in
restricted_or_unusable_edge
- Keep fake or overstated edge out of
usable_edge and unfair_advantages
Checklist
Examples
Example:
Input: I run a 20k newsletter for private practice accountants. People trust me,
but most readers are operators, not owners. I know their workflow pain well. I
also have internal templates from my employer, but I cannot reuse them.
Output: A profile that marks the newsletter as a real distribution asset but
buyer quality as mixed, records workflow knowledge as usable edge, marks the
employer templates as restricted edge, and asks follow-up questions about owner
reach and paid buyer access.
{
"user_profile": {
"skills": ["editorial packaging", "workflow analysis"],
"distribution": ["20k niche newsletter"],
"constraints": ["cannot reuse employer templates"]
},
"usable_edge": [
"Niche audience trust",
"Real workflow familiarity"
],
"restricted_or_unusable_edge": [
"Employer-owned templates"
],
"fake_or_overstated_edge": []
}
Common mistakes
| Mistake | Fix |
|---|
| Treating industry familiarity as a moat | Ask whether the user has real access, trust, or distribution |
| Treating employer access as usable edge | Move it to restricted_or_unusable_edge unless clearly usable |
| Treating follower count as buyer quality | Separate reach from monetizable trust and buyer access |
| Treating prestige or smart peers as moat | Put it under fake_or_overstated_edge unless it creates buyer leverage |
| Filling gaps with flattering assumptions | Name uncertainty and ask targeted follow-up |
Quick reference
| Operation | How |
|---|
| Extract profile | Pull concrete evidence across the core profile dimensions |
| Separate edge types | Split into usable, restricted, fake/overstated, and missing edge |
| Handle gaps | Ask short follow-ups or proceed with explicit assumptions |
| Format result | Human summary plus JSON using assets/profile-template.json |
Key principles
- Evidence over vibes — Profile from what the user actually said, not from
optimistic projection.
- Restricted edge is not usable edge — Employer, legal, or confidentiality
limits must stay visible.
- Distribution quality matters — Reach without buyer trust is not the same
as monetizable access.
- Do not inflate — Prestige, introductions, and vague familiarity do not
become moat without evidence.
- Weak edge is a valid conclusion — Say it when true instead of forcing a
stronger story.
Optimization Notes
- Preserve the user's requested output shape exactly and do not substitute generic advice for concrete artifacts.
- Include exact commands, code structures, protocol fields, tags, parameters, file paths, or deliverable sections when the task asks for them.
- Make safety gates explicit before irreversible, destructive, externally visible, or compliance-sensitive actions.
- For multi-step work, present steps in execution order and include validation or rollback checks where relevant.
- Avoid overfitting to a single eval example: express lessons as reusable rules, not as task-specific answers.