launch-validation-team — independently scanned and version-tracked by SaferSkills.
SaferSkills independently audited launch-validation-team (Agent Skill) and scored it 100/100 (green). The audit ran 55 deterministic rules across Security, Supply Chain, Maintenance, Transparency, and Community; it found 0 high-severity and 0 lower-severity findings. The full rule-by-rule trace and per-finding evidence are below. Free, methodology-open.
Findings & checks · 0 flagged
Every scanned point with the score it earned and what moved between them.
First recorded scan — no prior version to compare against.
The primary manifest — the file an agent reads to learn what this artifact does.
Spawn a standardized multi-agent team to validate a concept. Runs PM + HR + Specialists + Devil's Advocate in parallel. Produces a comprehensive viability report.
User says: "launch the team", "lanza el equipo", "validate this with agents", "run the full validation", "deep dive X", "is this real or hype".
Validated against: Tesla Fleet (killed), Option B tweet (verified + rewritten), CARRIERaaS (CAUTION with 3 paths), CITYaaS (fact-checked + mayor-ready). The pattern works. This skill makes it reusable.
Launch these agents IN PARALLEL (all in a single message with multiple tool calls):
Prompt template:
You are the PROJECT MANAGER. You DO NOT execute research. Your job is to organize and define the work for other specialized agents.
Mission: Validate [CONCEPT NAME].
Source files: [list relevant files]
Deliverables:
1. Research plan — 8-12 specific questions
2. Task assignments — which specialist owns which question
3. Critical path — the 3 questions whose NO answers kill the concept
4. Risk register — top 5 risks with severity
5. Scope discipline — what you are NOT doing
DO NOT execute research. DO NOT validate numbers. Organize only.Prompt template:
You are the HR / RECRUITING SPECIALIST. You DO NOT execute research. Your job is to define EXACT expert profiles needed to validate a complex concept.
Mission: [CONCEPT NAME]
Deliverables:
1. Specialist roster — 6-8 profiles needed with role, expertise, deliverable, anti-profile
2. Recruiting brief — 100-word "job description" per role
3. Team composition logic — why this mix covers the validation surface
4. Devil's advocate slot — one specialist whose job is to argue against
Domains to consider: Tech feasibility, legal, competitive, economics, data quality, partnership, buyer demand.Launch as many as needed (typically 3-5) based on the concept's risk profile. Each specialist has:
Common specialists:
Tech Reality Auditor — Reads API docs, SDK source, contracts, ToS. Verifies technical premises with primary sources.
Legal Scanner — Identifies required licenses, regulatory risk, cross-border issues. Produces jurisdiction matrix.
Competitor Mapper — Finds closest 3-5 competitors. Analyzes positioning, moats, failure modes.
Unit Economist — Builds bottom-up model with real buyer WTP + real worker cost. Produces sensitivity table.
Data Quality Analyst — Can evidence be produced, verified, legally admissible?
Partnership Scout — Finds 3-5 named first-customer targets with entry points.
Regulatory Strategist — Heat map of jurisdictions, insurance stack, legal framing. Useful for anything that crosses borders or touches regulated industries.
Prompt template:
You are the DEVIL'S ADVOCATE. Every other agent is trying to validate this concept. Your job is to argue AGAINST it as forcefully as possible. Find every reason it WON'T work.
Source files: [list]
Cover:
1. Why [gatekeeper] won't do this
2. Why [users] won't care
3. Why the data/product isn't valuable
4. Why the math doesn't work
5. Why [competitors] failed at this
6. Regulatory landmines
7. The real reason this won't get adopted
End with a verdict: VIABLE / VIABLE WITH CHANGES / NOT VIABLE.
Be brutal. Don't soft-pedal. The user explicitly wants ruthless critique.Understand what's being validated. Get the pitch, source docs, and user's specific concerns.
Use a single message with multiple Agent tool calls. All agents run concurrently.
Background mode is REQUIRED for validation teams. run_in_background: true on each agent. This prevents blocking the conversation.
As each agent completes, it notifies you. Read the output. Do NOT consolidate until all agents have returned.
Combine outputs into a single recommendation:
# [Concept Name] Validation Report
## The Team
- PM: [summary]
- HR: [team composition]
- Specialists: [list of roles]
- Devil's Advocate: [summary]
## The 3 Kill-Switches
1. [question] — [Y/N/UNKNOWN] — [evidence]
2. [question] — [Y/N/UNKNOWN] — [evidence]
3. [question] — [Y/N/UNKNOWN] — [evidence]
## 6-Axis Verdict
- A1 Technical Reality: [PASS/CAUTION/KILL]
- A2 Legal: [PASS/CAUTION/KILL]
- A3 Data Quality: [PASS/CAUTION/KILL]
- A4 Economics: [PASS/CAUTION/KILL]
- A5 Partnership: [PASS/CAUTION/KILL]
- A6 Buyer Demand: [PASS/CAUTION/KILL]
## The Brutal Summary
[One-paragraph verdict synthesizing all agents]
## 3 Paths Forward (if CAUTION)
1. [narrower scope]
2. [different customer]
3. [different jurisdiction/architecture]
## Kill List (if KILL)
- What to remove from the concept
- What to archive to graveyard
## Next Actions
- [concrete action with owner + deadline]Save the full report to vault/19-validation/scorecards/[concept-slug]-validation.md with proper frontmatter.
Run the validate-concept skill on the concept using the team's findings to produce the 1-page scorecard.
Each agent consumes context. Don't launch 10 agents when 5 would do.
~30 seconds. Free. No account. Every finding cites a rule and a line of evidence.