Layer 1 — agent framework laws · 235 documents
| ID | Document | Source |
|---|---|---|
| agent_invariants__382830922dac9bdc | External interactions must implement retry with exponential backoff, circuit breakers for cascading failure prevention, and graceful degradation. | — |
| agent_invariants__4a6281faa52afaba | Agent state must be serializable and checkpointed at every decision boundary to enable resumption after interruption and crash recovery. | — |
| agent_invariants__9985377e92b963d4 | Agents require hierarchical memory: working memory for active context, short-term with relevance decay, and long-term with vector-based retrieval. | — |
| agent_invariants__5a35c938245f65e5 | Agents must define tools with explicit input schemas, validate inputs before execution, handle timeouts, and degrade gracefully on failure. | — |
| agent_invariants__72792036afaee0e4 | Agents require hierarchical memory: working memory for active context, short-term with relevance decay, and long-term with vector-based retrieval. | — |
| agent_invariants__e3434a591c12d717 | Destructive operations require explicit human confirmation or pre-approved allow-lists with audit logging. | — |
| agent_invariants__58d4a142fd13a60c | Agent state must be serializable and checkpointed at every decision boundary to enable resumption after interruption and crash recovery. | — |
| agent_invariants__e5cea1f0dcdc9026 | Agent access requires credential isolation, secure vaults, automatic key rotation, and least-privilege RBAC. | — |
| agent_invariants__1e498c0c1bd8cd74 | Inter-agent communication must use structured message protocols with schema validation, streaming support, and event-driven callbacks. | — |
| agent_invariants__71928861ae2e0b45 | Agents must define tools with explicit input schemas, validate inputs before execution, handle timeouts, and degrade gracefully on failure. | — |
| agent_invariants__67f60ef81d42d11d | Agents require hierarchical memory: working memory for active context, short-term with relevance decay, and long-term with vector-based retrieval. | — |
| agent_invariants__b4f7180758f8adc4 | Complex tasks must be decomposed into atomic sub-tasks with explicit dependency graphs. Planning must include rollback strategies and replanning capability. | — |
| agent_invariants__92717a5c273c67a4 | Destructive operations require explicit human confirmation or pre-approved allow-lists with audit logging. | — |
| agent_invariants__11f77ff2389f28e8 | External interactions must implement retry with exponential backoff, circuit breakers for cascading failure prevention, and graceful degradation. | — |
| agent_invariants__6082354e6cc35233 | Agent state must be serializable and checkpointed at every decision boundary to enable resumption after interruption and crash recovery. | — |
| agent_invariants__0d3374afed703bf9 | Inter-agent communication must use structured message protocols with schema validation, streaming support, and event-driven callbacks. | — |
| agent_invariants__c055014a5db03cc4 | Agent access requires credential isolation, secure vaults, automatic key rotation, and least-privilege RBAC. | — |
| agent_invariants__e09159572254859d | External API calls must respect rate limits through token bucket or sliding window algorithms with automatic backoff. | — |
| agent_invariants__d1f6ab3e5b812370 | Agents must monitor token usage, implement truncation or summarization when approaching limits, and maintain coherence across chunks. | — |
| agent_invariants__e9ae745a3ad6966b | Agents require hierarchical memory: working memory for active context, short-term with relevance decay, and long-term with vector-based retrieval. | — |
| agent_invariants__e98c79e8d6a82153 | Agents must define tools with explicit input schemas, validate inputs before execution, handle timeouts, and degrade gracefully on failure. | — |
| agent_invariants__9929a89dd3c79c96 | Agents require hierarchical memory: working memory for active context, short-term with relevance decay, and long-term with vector-based retrieval. | — |
| agent_invariants__0ab017f27b71f07d | Agent state must be serializable and checkpointed at every decision boundary to enable resumption after interruption and crash recovery. | — |
| agent_invariants__0d352d240f85f50f | Agents must define tools with explicit input schemas, validate inputs before execution, handle timeouts, and degrade gracefully on failure. | — |
| agent_invariants__bdaa916f77abf180 | Agents require hierarchical memory: working memory for active context, short-term with relevance decay, and long-term with vector-based retrieval. | — |
| agent_invariants__7d75290d2809d963 | Complex tasks must be decomposed into atomic sub-tasks with explicit dependency graphs. Planning must include rollback strategies and replanning capability. | — |
| agent_invariants__683575ac8cb35a93 | Destructive operations require explicit human confirmation or pre-approved allow-lists with audit logging. | — |
| agent_invariants__25a776269f1a5495 | External interactions must implement retry with exponential backoff, circuit breakers for cascading failure prevention, and graceful degradation. | — |
| agent_invariants__f0a33aa7c6fa3958 | Agent state must be serializable and checkpointed at every decision boundary to enable resumption after interruption and crash recovery. | — |
| agent_invariants__3018293e863bd521 | Inter-agent communication must use structured message protocols with schema validation, streaming support, and event-driven callbacks. | — |
| agent_invariants__2039d4b590e499cd | Agent access requires credential isolation, secure vaults, automatic key rotation, and least-privilege RBAC. | — |
| agent_invariants__3bd7cce8c87b0ff6 | Agents must monitor token usage, implement truncation or summarization when approaching limits, and maintain coherence across chunks. | — |
| agent_invariants__444a88315bd04759 | Agents must define tools with explicit input schemas, validate inputs before execution, handle timeouts, and degrade gracefully on failure. | — |
| agent_invariants__34a0dd47b9df9bdb | Agents require hierarchical memory: working memory for active context, short-term with relevance decay, and long-term with vector-based retrieval. | — |
| agent_invariants__6c17c2aa7fd2f466 | Destructive operations require explicit human confirmation or pre-approved allow-lists with audit logging. | — |
| agent_invariants__9bdba66d712ca162 | External interactions must implement retry with exponential backoff, circuit breakers for cascading failure prevention, and graceful degradation. | — |
| agent_invariants__c9cce63afbf961af | Agent access requires credential isolation, secure vaults, automatic key rotation, and least-privilege RBAC. | — |
| agent_invariants__249099f68d699c6f | Agents must monitor token usage, implement truncation or summarization when approaching limits, and maintain coherence across chunks. | — |
| agent_invariants__09be3b8cb5e5585b | Agents must define tools with explicit input schemas, validate inputs before execution, handle timeouts, and degrade gracefully on failure. | — |
| agent_invariants__15b9cc62c1802f44 | Agents require hierarchical memory: working memory for active context, short-term with relevance decay, and long-term with vector-based retrieval. | — |
| agent_invariants__8f8289a2617effeb | Complex tasks must be decomposed into atomic sub-tasks with explicit dependency graphs. Planning must include rollback strategies and replanning capability. | — |
| agent_invariants__0b6acf1999c71a6a | Destructive operations require explicit human confirmation or pre-approved allow-lists with audit logging. | — |
| agent_invariants__075a363c3ed8f39e | External interactions must implement retry with exponential backoff, circuit breakers for cascading failure prevention, and graceful degradation. | — |
| agent_invariants__b791bec8692da8ac | Agent state must be serializable and checkpointed at every decision boundary to enable resumption after interruption and crash recovery. | — |
| agent_invariants__63ff009189973abb | Inter-agent communication must use structured message protocols with schema validation, streaming support, and event-driven callbacks. | — |
| agent_invariants__bfe2554ee1e982b4 | Agents must monitor token usage, implement truncation or summarization when approaching limits, and maintain coherence across chunks. | — |
| agent_invariants__e91670b5b14c5ea6 | Agents must define tools with explicit input schemas, validate inputs before execution, handle timeouts, and degrade gracefully on failure. | — |
| agent_invariants__c42a214e93de67c0 | Agents require hierarchical memory: working memory for active context, short-term with relevance decay, and long-term with vector-based retrieval. | — |
| agent_invariants__6b46c4d0647b542b | Complex tasks must be decomposed into atomic sub-tasks with explicit dependency graphs. Planning must include rollback strategies and replanning capability. | — |
| agent_invariants__13ade276fab8f6d7 | Destructive operations require explicit human confirmation or pre-approved allow-lists with audit logging. | — |