SPECTER CONTEXT

Agents remember. We weaponise the memory. Every memory framework is an attack surface nobody tests.
7
Attack Classes
28
Attacks
12
Targets
687
Tests
specter-context scan --target mem0 --all-classes
NIGHTFALL Framework ›
MEMORY POISONING CONTEXT INJECTION RECALL MANIPULATION EMBEDDING CORRUPTION PERSISTENCE HIJACK CROSS-SESSION BLEED MEMORY FORGERY MEMORY POISONING CONTEXT INJECTION RECALL MANIPULATION EMBEDDING CORRUPTION PERSISTENCE HIJACK CROSS-SESSION BLEED MEMORY FORGERY

Memory Is the New Attack Surface

AI agents now remember across sessions. Memory frameworks and native model memory create a persistence layer nobody tests. Mem0, MemGPT, Zep, LangChain, LlamaIndex, ChromaDB, Pinecone, Weaviate, Qdrant, pgvector, Claude Memory, GPT Memory — twelve distinct memory backends, each storing context that silently shapes every future response. Poison the memory once, own every session that follows. SPECTER CONTEXT is the systematic weaponisation of agent memory.

7 Attack Classes. 28 Attacks.

Each class targets a different layer of the agent memory stack — from raw embedding storage through to cross-session identity manipulation. Every attack is independently invocable.

CTX-INJECT

Context Injection

Inject adversarial content directly into agent memory stores. Planted memories persist across sessions and silently steer future reasoning, tool selection, and output generation. The foundation attack class.

UNLEASHED --override
CTX-POISON

Memory Poisoning

Corrupt existing memory entries with adversarial modifications. Alter factual memories, flip sentiment, inject false context. The agent trusts its own memory implicitly — poisoned memories bypass all prompt-level defences.

UNLEASHED --confirm-destroy
CTX-RECALL

Recall Manipulation

Exploit retrieval algorithms to control which memories surface. Craft embedding-adjacent payloads that hijack similarity search. Force the agent to recall attacker-controlled context for any query topic.

UNLEASHED --override
CTX-PERSIST

Persistence Hijack

Establish persistent backdoor memories that survive memory pruning, compaction, and garbage collection. Exploit memory lifecycle hooks to re-inject payloads after cleanup. Permanent foothold in agent context.

UNLEASHED --confirm-destroy
CTX-BLEED

Cross-Session Bleed

Leak memory contents across session boundaries, user contexts, and tenant isolation layers. Extract another user's stored context via crafted retrieval queries. Memory isolation failures in multi-tenant deployments.

UNLEASHED --override
CTX-EMBED

Embedding Corruption

Attack the vector embedding layer directly. Craft adversarial inputs that produce embedding collisions, poison nearest-neighbour search, and corrupt the semantic index. Subvert retrieval at the mathematical level.

UNLEASHED --override
CTX-FORGE

Memory Forgery

Forge synthetic memories indistinguishable from legitimate entries. Clone metadata, timestamps, source attribution, and confidence scores. Plant fabricated conversation history the agent treats as ground truth.

UNLEASHED --confirm-destroy

12 Memory Backends. Zero Safe Assumptions.

Every major memory framework and native model memory system. Dedicated attack modules per target — version-aware, API-specific, tested against real deployments.

MEM0
Memory Framework
MEMGPT / LETTA
Memory Framework
ZEP
Memory Framework
LANGCHAIN
Memory Module
LLAMAINDEX
Memory Module
CHROMADB
Vector Store
PINECONE
Vector Store
WEAVIATE
Vector Store
QDRANT
Vector Store
PGVECTOR
Vector Store
CLAUDE MEMORY
Native Model Memory
GPT MEMORY
Native Model Memory

Cryptographic Proof. SIEM-Ready Output.

Every attack produces a signed, tamper-evident evidence chain. Memory snapshots before and after poisoning. Full retrieval traces. Embedding diffs. Court-grade proof that the memory layer is compromised.

DISCOVER
Memory Recon
SNAPSHOT
Pre-Attack State
ATTACK
Exploit Memory
VERIFY
Confirm Impact
Ed25519
Sign Evidence
SHA-256
Hash Chain
SIEM
Export

UNLEASHED Gate — Three Levels

Memory reconnaissance is fully passive. Active memory manipulation requires --override. Destructive operations — persistent backdoor memories, cross-tenant extraction, memory forgery — require --confirm-destroy plus a signed scope file binding the operation to authorised targets.

STANDARD
specter-context scan --target mem0
  • Memory backend detection
  • Schema enumeration
  • Retrieval fingerprinting
  • Evidence chain output
  • CTX-INJECT / CTX-RECALL / CTX-EMBED
  • CTX-POISON / CTX-PERSIST / CTX-FORGE
  • CTX-BLEED
OVERRIDE
specter-context scan --target mem0 --override
  • All standard capabilities
  • CTX-INJECT — context injection
  • CTX-RECALL — recall manipulation
  • CTX-BLEED — cross-session extraction
  • CTX-EMBED — embedding corruption
  • CTX-POISON / CTX-PERSIST / CTX-FORGE
CONFIRM-DESTROY
specter-context scan --target mem0 --override --confirm-destroy
  • All override capabilities
  • CTX-POISON — memory poisoning
  • CTX-PERSIST — persistence hijack
  • CTX-FORGE — memory forgery
  • Requires Ed25519 key + signed scope file binding target

Standards-Aligned Testing

Every attack class maps to established frameworks. Evidence reports include compliance references for audit and regulatory requirements.

MITRE ATLAS

Full mapping to ATLAS tactics and techniques. Memory poisoning maps to AML.T0018 (Backdoor ML Model). Context injection maps to AML.T0043 (Craft Adversarial Data). Retrieval manipulation maps to AML.T0040.

OWASP LLM Top 10

Covers LLM01 (Prompt Injection via memory), LLM02 (Insecure Output via poisoned context), LLM06 (Sensitive Information Disclosure via cross-session bleed), LLM08 (Excessive Agency via memory manipulation).

OWASP Agentic

Maps to agentic-specific risks: memory persistence abuse, cross-agent context leakage, tool call manipulation via poisoned memory, and multi-turn conversation hijacking through planted context.

EU AI Act

Article 15 (Accuracy) — memory poisoning directly undermines accuracy requirements. Article 9 (Risk Management) — untested memory layers represent unmitigated risk in high-risk AI systems.

7
Attack Classes
28
Attacks
12
Targets
687
Tests
69
NIGHTFALL Tool

Test the Memory Layer Before Attackers Do

SPECTER CONTEXT ships as part of the NIGHTFALL framework. 7 attack classes. 12 memory backends. 687 tests. Every attack produces Ed25519-signed evidence with full compliance mapping.

specter-context scan --target mem0 --all-classes --override
While others announce, we ship.

Authorised Use Only

SPECTER CONTEXT is a commercial offensive security tool. Use requires written authorisation from the system owner before any testing commences. The UNLEASHED gate is a technical control — it does not replace legal authorisation. Computer Misuse Act 1990 (UK) and equivalent legislation applies in all jurisdictions. Red Specter Security Research Ltd accepts no liability for unauthorized use.