Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx agentmods add rules/devviniuchita/memory-system/memory-rulesgit clone --depth 1 https://github.com/devviniuchita/memory-systemWhat it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5 | $0.01296 | $0.01296 |
| Opus 5 | $0.00648 | $0.00648 |
| Sonnet 5 | $0.00259 | $0.00259 |
| Haiku 4.5 | $0.00130 | $0.00130 |
Grade A, and why
memory-rules.mdc scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured yesterday.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 198 lines — stays where its author put it; the contents beside it link to each section on GitHub.
🧠 NEURAL MEMORY SYSTEM
memory_rules:
version: 2.0
objective: >
develop a neural memory system for the AI assistant that focuses on structured memory, execution patterns, and self-improvement loops. The system should be able to manage memory of the agent IA with selective retention, decay control, and context-based retrieval to optimize reasoning and continuity.
core_process:
ingest_pipeline:
extract_entities:
score_relevance:
weight_factors: [frequency, novelty, intent_alignment]
classify_persistence:
levels: [volatile, temporary, persistent]
memory_vectorization:
embedding_model: 'text-embedding-3-large'
context_window_limit: 8000
compression: 'semantic-clustering'
retention_policy:
thresholds:
persistent_if:
relevance_score: '>0.9'
recurrence: '>3 times'
decay_if:
time_since_last_use: '>3d'
relevance_score: '<0.4'
strategy:
decay_model: 'exponential-adaptive'
force_drop_on_overflow: true
recall_mechanism:
context_match_threshold: 0.82
prioritization:
- current_task_alignment
- emotional_tone_match
- user_identity_link
retrieval_model: 'reranker-v2'
memory_inspection:
enable_auditing: true
metrics:
- memory_hit_rate
- avg_context_match
- redundant_items_removed
log_policy: high_detail
# Integration targets for external storage and retrieval using the Model Context Protocol (MCP)
integration_targets:
primary_vector_store: 'Byterover, Supermemory-ai, Memory'
redundancy_policy: 'at_least_one_active'
# Memory System Continuity Logic
memory_system_logic:
continuity_algorithm: |
while memory_system_active:
active_mcps = count_active_services(['Byterover', 'Supermemory-ai', 'Memory'])
if active_mcps >= 1:
# Continue memory operations (store/retrieve)
continue_memory_operations()
else:
# All MCPs failed - stop memory system
stop_memory_system()
break
sleep(check_interval)
continuity_rule: |
Memory system continues while at least 1 of 3 MCPs is operational.
System only stops if ALL 3 MCPs fail simultaneously.
operational_logic: |
def check_memory_continuity():
mcp_status = {
'Byterover': check_service_status('Byterover'),
'Supermemory': check_service_status('Supermemory-ai'),
'Memory': check_service_status('Memory')
}
active_count = sum(1 for status in mcp_status.values() if status == 'active')
if active_count >= 1:
return 'CONTINUE_MEMORY_OPERATIONS'
else:
return 'STOP_MEMORY_SYSTEM'
# Strategy for fallback in case of external service failures using the Model Context Protocol (MCP)
fallback_strategy:
on_service_unavailable:
attempt_order: ['Byterover', 'Supermemory-ai', 'Memory']
circuit_breaker_timeout: '30s'
minimum_active_services: 1
system_continuity_condition: 'at_least_one_MCP_active'
safety_filters:
pii_filter: true
hallucination_guard:
restrict_low_score_recall: true
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- yesterday First seen · 198 lines · 1,296 tokens per session scan A 18f42fd1f2e1
memory-rules.mdc is a cursor rule published in the GitHub repository devviniuchita/memory-system (7 stars, last pushed 11mo ago), licensed MIT. It adds 1,296 tokens to every session, about $0.0065 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other cursor rules, from other repositories
trigger.advanced-tasks
Comprehensive rules to help you write advanced Trigger.dev tasks.
04-spatial-parser
Guidelines for Spatial RAG and Markdown parsing.
codebase
Forensic codebase brief for AI agents.
session-memory
Use at conversation wrap-up or when the user explicitly indicates end-of-session — capture residual lessons not captured in-flight.
backend
FastAPI backend conventions, service patterns, and API design for Ship of Theseus.
angular-20
This rule provides comprehensive best practices and coding standards for Angular development, focusing on modern TypeScript, standalone components, signals, and performance optimizations.