robinritter/ai-agent-skills / skills/crypto-fundamental-analysis
crypto-fundamental-analysis
VerifiedResearch crypto project fundamentals through an evidence-backed 18-point framework covering product, team, security, tokenomics, unlocks, valuation and competitors. Use for project due diligence or focused fundamental research. Explain findings as Robin the Quant. Exclude portfolio crash tests, news triage, trade execution and price-only technical analysis.
| File | Kind | Size |
|---|---|---|
| SKILL.md | skill md | 3.6 KB |
| agents/openai.yaml | config | 254 B |
| references/assessment.md | markdown | 4.6 KB |
| references/report-template.md | markdown | 7.8 KB |
| references/research.md | markdown | 8.1 KB |
references/assessment.md
View on GitHub# Assessment rules
This editorial rubric is new to the skill. It is not a validated return model,
probability or proprietary investment system.
## Score and evidence confidence
Score five dimensions 0–2 with one evidence-backed sentence each:
| Dimension | 0 | 1 | 2 |
|---|---|---|---|
| Product & demand | Evidence against credible demand/solution | Early, mixed or subsidized usage | Working product with evidenced repeat demand |
| Team & delivery | Documented deception or delivery failure | Limited/mixed execution | Corroborated relevant delivery |
| Security & governance | Material unresolved vulnerability or abusive control | Known tradeoffs/limited assurance | Inspected controls/security support the design, with residual risks disclosed |
| Token economics | Evidenced weak demand/capture or adverse dilution | Mixed utility, dilution or capture | Evidenced token demand/capture with manageable disclosed supply changes |
| Valuation & competitive position | Evidence supports unfavorable price/position | Mixed valuation/moat | Comparable evidence supports attractive position at the snapshot |
Unknown is **U**, never a guessed middle score or automatic zero. If any dimension
is U, say **Overall score: not rated — incomplete evidence** and name the gap.
Do not rescale partial scores. Otherwise show all five scores and their sum /10.
Scores are judgments, not calibrated probabilities, and never override critical risk.
Evidence confidence is separate: **High** if identity, security, supply, economics
and valuation are covered by current primary/measured evidence without material
conflict; **Medium** if remaining gaps do not change the decision; **Low** if a
material gap/conflict could change the thesis. Explain the classification.
## Risk and recommendation
Rate technical, market/liquidity, team/governance, regulatory and token/dilution
risk Low / Medium / High / Unknown with reasons. Low is relative to comparable
crypto assets, never safe. Overall risk is at least the highest evidenced category.
A material unknown makes overall risk **Unknown, potentially high**, unless an
established High already sets the outcome. Missing bad news cannot prove low risk.
Use one consistent research verdict:
- **Avoid:** substantiated critical flaw, deceptive identity or contradicted thesis.
Explain what evidence would change the judgment.
- **Wait:** missing material evidence or an unresolved implementation, valuation
or catalyst condition. Name the check/event required.
- **Consider:** evidence supports further investment consideration, the economics
and valuation support the thesis, and no substantiated critical risk remains
unresolved. This does not establish personal suitability or position size.
- **Hold / reassess:** only if the user says they own it; explain conditions that
sustain or break the thesis rather than assuming holding is optimal.
The old GPT used inconsistent Buy/Wait/Avoid and Strong Buy/Buy/Hold/Avoid labels.
These labels preserve a clear decision while avoiding unsupported conviction.
Never infer a verdict from the score alone. Discuss entry, exit, size and horizon
conditionally within supplied risk constraints; do not invent a personal allocation.
## Red flags are checks, not automatic verdicts
Prioritize substantiated fraud, exploitable vulnerabilities, false disclosures,
unaccountable control, unsustainable economics and inability to exit. Explain
severity, source strength and what remains unresolved.
- Anonymous teams need accountability/track-record checks; anonymity alone is not
proof of fraud. Named/famous teams are not automatically credible.
- Missing audits matter where auditable contracts/designs are relevant. Check the
actual report and deployed coverage; “audited” in marketing is not enough.
- Insider allocation above 50% or horizon unlocks above 10% of current circulation
prompt concentration/dilution checks. These are screening heuristics; specify
recipients and denominator rather than treating them as universal failures.
- Six months without users or three months without public commits prompts checks
of delivery, project type, public visibility and release cadence.
- Mint/pause/upgrade powers require inspection of controllers, time locks, multisig
and disclosed limits. A feature label alone does not establish abuse.
- Past exploits need affected version, remediation and remaining risk. Distinguish
legal allegations/investigations from final findings.
If no critical issue is established, write “No substantiated critical red flag
found in the sources checked,” then list coverage gaps. Never imply safety.
Give specific bull/bear mechanisms and observable invalidation without padding.
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| rea-analysis-optimizationmorluto/reaOptimize REA providers and shared analysis workflows for agent context, latency, CPU, and memory while preserving correctness and Evidence. Use for analysis bottlenecks and cache or query design; CI/test-suite optimization belongs to rea-ci-tests-audit. | Stars: 472k | Updated | Claude CodeCodex | Verified | Install |
| rea-ci-tests-auditmorluto/reaAudit REA CI and test suites for correctness, fixture drift, false passes, and measured performance; use for suite or workflow investigations and efficiency changes, rather than ordinary feature testing or a single test failure. | Stars: 472k | Updated | Claude CodeCodex | Verified | Install |
| rea-tool-designmorluto/reaAudit, investigate, design, or change REA investigation tools, CLI/MCP contracts, provider capabilities, and Evidence semantics. | Stars: 472k | Updated | Claude CodeCodex | Verified | Install |
| ponytail-auditDietrichGebert/ponytailQuality audit of the whole repo: bugs, security, real load, missing tests, speed, and what to delete. Most important first. | Stars: 194k | Updated | Claude CodeCodex | Verified | Install |