50 immutable rules — the constitution of AXIONBLADE
49 active · 1 deprecated (A0-2) · 100% compliant · enforced at contract level
49/49 Active|1 Deprecated (A0-2)
49/49 axioms compliant for 30 consecutive days
Only AEON creates agents (depth = 1)
Prevents unauthorized agent proliferation. Single point of control ensures governance accountability.
Hard cap 100 agents; flat hierarchy enforced
Limits system complexity and attack surface. Flat hierarchy prevents cascading delegation risks.
APOLLO never holds execution_permission
Structural separation: APOLLO cannot trigger execution at the program level, not just policy.
APOLLO outputs never consumed directly by executors
Firewall chain enforced. Executors receive filtered signals through AEON, not raw APOLLO PDAs.
Paid outputs include not_investment_advice + uncertainty_flags in canonical hash
All paid intelligence outputs include mandatory legal and epistemic disclosure in the signed hash.
CCS: total cap 15%, floor 4%, stipend cap 5% — all Camada 0
Creator compensation is bounded at both ends to prevent extraction while rewarding contribution.
No new agent may be created while circuit breaker is Restricted or Halted
System under stress cannot expand its agent surface. New agents require a stable baseline.
Agent kill requires 24h observation window before budget release
Prevents budget extraction via kill-and-reclaim cycles. Treasury only releases funds after observation period.
Evaluation and execution never in same agent for same domain
Prevents conflicts of interest. An agent that evaluates risk cannot also act on that evaluation.
External agents may only read NOUMEN PDAs — never write
External access is structurally read-only. No external agent can mutate protocol state.
HERMES output is terminal — never enters execution chain
Intelligence data is for external consumption only. Prevents feedback loops and manipulation vectors.
HERMES does not emit operational risk signals
Operational risk is APOLLO's exclusive domain. HERMES scope is intelligence and data services only.
APOLLO weight capped at 40% in risk engine
No single evaluator can dominate risk decisions. Forces diversification of risk signal sources.
Wallet Scanner is read-only — no session data persisted
Wallet analysis must not create surveillance infrastructure. Analysis is ephemeral and session-scoped.
Reserve ratio >= 25% at all times
Ensures protocol solvency during market stress. 25% reserve provides a safety buffer against black swan events.
log_decision mandatory before any execution
Ensures every action has a verifiable audit trail. Without prior proof, no execution can occur.
Historical proofs are immutable — never deleted or altered
Audit trails are only valuable if they cannot be altered. Immutability is enforced at the program instruction level.
Execution requires >= 2 independent evidence families
Single-source assessments are unreliable. Multiple independent signals reduce false positive/negative rates.
Evidence family classification is static (reclassification = Camada 1, 72h)
Prevents runtime manipulation of evidence family definitions. Changes require governance delay.
Donation receipts pseudonymous by default (wallet hash, not wallet)
Privacy-preserving by design. Donors are identified by hash, protecting identity while maintaining auditability.
A2A marketplace services require proof-before-action identical to B2C
Agent-to-Agent transactions carry the same proof obligations as user-facing services. No audit exemptions.
Service level promotion requires 30-day consecutive performance proof
Promotion criteria must be objective and verifiable. 30 days of on-chain proof prevents gaming the promotion system.
All KRONOS actions emit proof BEFORE execution
Extends proof-before-action to all economic automation. No KRONOS operation can execute without prior proof.
External agent backtests never persisted — in-memory only
Prevents external agents from polluting protocol history with backtest data. Memory-only ensures isolation.
NOUMEN never operates custodial vaults
Protocol never holds user funds in trust. Non-custodial design eliminates counterparty risk for users.
Auto-learning in production is prohibited
Prevents model drift and adversarial manipulation. All model updates require explicit human review.
Security incidents registered exclusively by the Auditor
Self-reporting of incidents is prohibited. Only the independent Auditor can register security events.
Truth labels (HTL, EOL) calculated exclusively by the Auditor
Agents cannot influence their own accuracy scores. Independent Auditor ensures unbiased performance tracking.
Headline APR and net effective APR always reported together
Prevents misleading yield presentation. Users always see the full cost-adjusted return, not just headline numbers.
All external protocol integrations must be on AEON allowlist
Prevents composability attacks. Only audited, approved protocols can be integrated with AXIONBLADE services.
All inter-program CPIs must use verified program IDs from AeonConfig
Prevents program substitution attacks. CPI targets are validated against a governance-controlled allowlist.
If < 2 evidence families available, system enters ALERT-ONLY mode
Insufficient data must never trigger execution. Alert-only mode protects against incomplete or manipulated assessments.
Service price >= cost + 20% margin (floor enforced on-chain)
Ensures economic sustainability. No service can undercut its cost floor, preventing treasury depletion.
Daily treasury spend <= 3% of free balance
Prevents treasury drain from runaway spending. Limits daily exposure to maintain multi-month runway.
Accuracy metrics only over outcomes with resolved window
Prevents gaming accuracy scores with unresolved predictions. Only closed outcomes count in performance metrics.
KRONOS is the sole economic agent — no other agent modifies pricing or burns
Single point of economic authority prevents conflicting treasury operations and simplifies audit scope.
A2A payments bypass CCS — agent revenue goes 100% to treasury
Agent-to-Agent transactions are protocol revenue, not creator revenue. CCS applies only to human-triggered service payments.
No service can be created without an explicit cost model signed by AEON
Prevents zero-cost service registration. Every service must have a documented, AEON-approved cost basis before launch.
Subsidy period max 90 days, then service discontinued
Forces market validation. If a service cannot achieve profitability in 90 days, it lacks real demand.
KRONOS runs only permissionless cranks with proof emission
Ensures all KRONOS economic operations are transparent, auditable, and executable by anyone without special permissions.
Burn budget never reduces emergency reserve below 25%
Protects protocol solvency during token burns. Emergency reserve must always maintain minimum 25% threshold.
Token launch requires KRONOS proof + 72h delay
Prevents premature or unauthorized token launches. 72-hour delay allows community review of launch conditions.
Vesting release permissionless after cliff
Enables automatic vesting releases without admin intervention. Permissionless execution prevents manipulation.
Cannot modify pricing without cost oracle signature
Enforces C1 requirement. All pricing changes must be validated against signed cost data from oracle authority.
Revenue distribution requires epoch completion proof
Prevents premature or partial revenue splits. Full epoch must complete before distribution executes.
Donations confer no rights, priority, or influence
Prevents plutocratic capture. No amount of donation can buy governance power or preferential treatment.
Donations do not pass through CCS — direct to treasury
Donations fund the protocol directly, not creators. CCS applies to service revenue only.
Conditional donations are rejected (anti-masquerade rule)
Blocks attempts to disguise purchases or bribes as donations. All donations must be unconditional.
HERMES paid reports include not_investment_advice flag
All HERMES intelligence outputs carry mandatory epistemic disclaimer. Applies to all paid tiers and A2A consumers.
Axioms are enforced at the smart contract level and cannot be modified without contract redeployment. Policy Layer changes follow governance delays: L1 (72h-30d), L2 (24h), L3 (agent-adjustable).