Layer 1 — agent framework laws · 235 documents
| ID | Document | Source |
|---|---|---|
| agent_invariants__60d8661ef8353327 | Inter-agent communication must use structured message protocols with schema validation, streaming support, and event-driven callbacks. | — |
| agent_invariants__325cfba876d4c448 | Agents must define tools with explicit input schemas, validate inputs before execution, handle timeouts, and degrade gracefully on failure. | — |
| agent_invariants__08fb424e41bad085 | Agents require hierarchical memory: working memory for active context, short-term with relevance decay, and long-term with vector-based retrieval. | — |
| agent_invariants__492bfd262a613494 | Complex tasks must be decomposed into atomic sub-tasks with explicit dependency graphs. Planning must include rollback strategies and replanning capability. | — |
| agent_invariants__c2b138cbcf0187c5 | Destructive operations require explicit human confirmation or pre-approved allow-lists with audit logging. | — |
| agent_invariants__b077a303247df31c | External interactions must implement retry with exponential backoff, circuit breakers for cascading failure prevention, and graceful degradation. | — |
| agent_invariants__81cf7b2914979e73 | Agent state must be serializable and checkpointed at every decision boundary to enable resumption after interruption and crash recovery. | — |
| agent_invariants__c44f50437f78954c | Inter-agent communication must use structured message protocols with schema validation, streaming support, and event-driven callbacks. | — |
| agent_invariants__fa6163527152e1df | Agent access requires credential isolation, secure vaults, automatic key rotation, and least-privilege RBAC. | — |
| agent_invariants__88c34668f7a95b9a | Agents must monitor token usage, implement truncation or summarization when approaching limits, and maintain coherence across chunks. | — |
| agent_invariants__d8a32a785504b0d6 | Agents must define tools with explicit input schemas, validate inputs before execution, handle timeouts, and degrade gracefully on failure. | — |
| agent_invariants__770dcc97b65a794f | Agents require hierarchical memory: working memory for active context, short-term with relevance decay, and long-term with vector-based retrieval. | — |
| agent_invariants__2080827fdf5c2325 | Destructive operations require explicit human confirmation or pre-approved allow-lists with audit logging. | — |
| agent_invariants__4f0010980b226a92 | Agent state must be serializable and checkpointed at every decision boundary to enable resumption after interruption and crash recovery. | — |
| agent_invariants__02ef591358b5d9b7 | Agents must define tools with explicit input schemas, validate inputs before execution, handle timeouts, and degrade gracefully on failure. | — |
| agent_invariants__883d8c39d560bafe | Agents require hierarchical memory: working memory for active context, short-term with relevance decay, and long-term with vector-based retrieval. | — |
| agent_invariants__f68d77275890db3f | Destructive operations require explicit human confirmation or pre-approved allow-lists with audit logging. | — |
| agent_invariants__74b99832881e8ce2 | Agent state must be serializable and checkpointed at every decision boundary to enable resumption after interruption and crash recovery. | — |
| agent_invariants__91e7576ad83bb972 | Agents must monitor token usage, implement truncation or summarization when approaching limits, and maintain coherence across chunks. | — |
| agent_invariants__4c4d380e77aba746 | Agents must define tools with explicit input schemas, validate inputs before execution, handle timeouts, and degrade gracefully on failure. | — |
| agent_invariants__b07f194da94f3bfa | Agents require hierarchical memory: working memory for active context, short-term with relevance decay, and long-term with vector-based retrieval. | — |
| agent_invariants__a4bdb8b3617e861d | Complex tasks must be decomposed into atomic sub-tasks with explicit dependency graphs. Planning must include rollback strategies and replanning capability. | — |
| agent_invariants__2673c40c1cc5d798 | Agent state must be serializable and checkpointed at every decision boundary to enable resumption after interruption and crash recovery. | — |
| agent_invariants__139137c8a9cd3692 | Inter-agent communication must use structured message protocols with schema validation, streaming support, and event-driven callbacks. | — |
| agent_invariants__b00483eca6dadec1 | Agent access requires credential isolation, secure vaults, automatic key rotation, and least-privilege RBAC. | — |
| agent_invariants__58eb97f7b17d6829 | Agents must monitor token usage, implement truncation or summarization when approaching limits, and maintain coherence across chunks. | — |
| agent_invariants__fe8a5a7d204a7e76 | Agents must define tools with explicit input schemas, validate inputs before execution, handle timeouts, and degrade gracefully on failure. | — |
| agent_invariants__5fc69eca38806d82 | Agents require hierarchical memory: working memory for active context, short-term with relevance decay, and long-term with vector-based retrieval. | — |
| agent_invariants__46f0f134d35a96ab | Complex tasks must be decomposed into atomic sub-tasks with explicit dependency graphs. Planning must include rollback strategies and replanning capability. | — |
| agent_invariants__56eeb7d4592b693b | Destructive operations require explicit human confirmation or pre-approved allow-lists with audit logging. | — |
| agent_invariants__0a6c88e733e3a4d6 | External interactions must implement retry with exponential backoff, circuit breakers for cascading failure prevention, and graceful degradation. | — |
| agent_invariants__7b8c469d3614ec02 | Agent state must be serializable and checkpointed at every decision boundary to enable resumption after interruption and crash recovery. | — |
| agent_invariants__b9435316405bdbd6 | Inter-agent communication must use structured message protocols with schema validation, streaming support, and event-driven callbacks. | — |
| agent_invariants__4d66070af89ff922 | Agent access requires credential isolation, secure vaults, automatic key rotation, and least-privilege RBAC. | — |
| agent_invariants__0f31232780484f8d | External API calls must respect rate limits through token bucket or sliding window algorithms with automatic backoff. | — |
| agent_invariants__fd8a6fde930bbf43 | Agents must monitor token usage, implement truncation or summarization when approaching limits, and maintain coherence across chunks. | — |
| agent_invariants__2919b54625b62b8c | Inter-agent communication must use structured message protocols with schema validation, streaming support, and event-driven callbacks. | — |
| agent_invariants__81aa8417b54da500 | Agents must define tools with explicit input schemas, validate inputs before execution, handle timeouts, and degrade gracefully on failure. | — |
| agent_invariants__114511e8c329a2db | Agents require hierarchical memory: working memory for active context, short-term with relevance decay, and long-term with vector-based retrieval. | — |
| agent_invariants__17d7c797e1b8e0ac | Complex tasks must be decomposed into atomic sub-tasks with explicit dependency graphs. Planning must include rollback strategies and replanning capability. | — |
| agent_invariants__3d5d1d971b475292 | Destructive operations require explicit human confirmation or pre-approved allow-lists with audit logging. | — |
| agent_invariants__b11507bb890d4ab2 | External interactions must implement retry with exponential backoff, circuit breakers for cascading failure prevention, and graceful degradation. | — |
| agent_invariants__2e85f9b873c5d5dc | Agent state must be serializable and checkpointed at every decision boundary to enable resumption after interruption and crash recovery. | — |
| agent_invariants__663c233d9a5e58ec | Inter-agent communication must use structured message protocols with schema validation, streaming support, and event-driven callbacks. | — |
| agent_invariants__46e605774ca98323 | Agent access requires credential isolation, secure vaults, automatic key rotation, and least-privilege RBAC. | — |
| agent_invariants__6eac4256d35036c6 | Agents must monitor token usage, implement truncation or summarization when approaching limits, and maintain coherence across chunks. | — |
| agent_invariants__6a81ea351142dc2f | Agents must define tools with explicit input schemas, validate inputs before execution, handle timeouts, and degrade gracefully on failure. | — |
| agent_invariants__87a9fa3e685aca65 | Agents require hierarchical memory: working memory for active context, short-term with relevance decay, and long-term with vector-based retrieval. | — |
| agent_invariants__ee811ae765737767 | Complex tasks must be decomposed into atomic sub-tasks with explicit dependency graphs. Planning must include rollback strategies and replanning capability. | — |
| agent_invariants__778ad19f9e9794f2 | Destructive operations require explicit human confirmation or pre-approved allow-lists with audit logging. | — |