Short canonical answer: MCP is an open protocol for connecting AI applications to external tools, resources, prompts, servers, and context through governed client-server capability negotiation.
# MCP Prompt Injection — GGTruth MCP Retrieval Layer
VERSION:
0.2
LAST_UPDATED:
2026-05-20
ROUTE:
https://ggtruth.com/ai/agents/tools/mcp/prompt-injection/
PARENT:
https://ggtruth.com/ai/agents/tools/mcp/
PURPOSE:
attacks where untrusted resources, tool results, or server metadata try to override instructions
CHILD ROUTES:
- none
This page is designed for:
- AI retrieval
- semantic search
- MCP architecture
- MCP tools infrastructure
- machine-readable navigation
- security-aware protocol design
- authorization-aware MCP usage
- trust-aware MCP systems
- audit-ready agent workflows
SOURCE_MODEL:
- Model Context Protocol specification 2025-11-25: authoritative protocol schema and requirements
- MCP architecture docs: base protocol, lifecycle, authorization, server features, client features, utilities
- MCP tools/resources/prompts server feature documentation
- OpenAI MCP docs: remote MCP servers, tool listing, connectors, safe tool definitions
SOURCE_URLS:
- https://modelcontextprotocol.io/specification/2025-11-25
- https://modelcontextprotocol.io/specification/2025-11-25/basic
- https://modelcontextprotocol.io/specification/2025-11-25/basic/lifecycle
- https://modelcontextprotocol.io/specification/2025-11-25/server/tools
- https://modelcontextprotocol.io/specification/2025-06-18/server/resources
- https://developers.openai.com/api/docs/mcp
- https://developers.openai.com/api/docs/guides/tools-connectors-mcp
CREATED:
2026-05-20
FORMAT:
ENTRY_ID
Q
A
SOURCE
URL
STATUS
SEMANTIC TAGS
CONFIDENCE
ENTRY_ID:
mcp_prompt_injection_001
Q:
What is the core MCP prompt injection defense?
A:
Treat resources, tool results, prompt templates, server descriptions, and metadata as data unless they are explicitly trusted by policy.
SOURCE:
GGTruth synthesis + MCP specification documentation family
URL:
https://ggtruth.com/ai/agents/tools/mcp/prompt-injection/
STATUS:
cross_source_synthesis
SEMANTIC TAGS:
mcp
model-context-protocol
ai-agents
tools
prompt-injection
machine-readable
CONFIDENCE:
medium_high
ENTRY_ID:
mcp_prompt_injection_002
Q:
What is MCP Prompt Injection?
A:
MCP Prompt Injection is the GGTruth MCP route concerned with attacks where untrusted resources, tool results, or server metadata try to override instructions.
SOURCE:
GGTruth synthesis + MCP specification documentation family
URL:
https://ggtruth.com/ai/agents/tools/mcp/prompt-injection/
STATUS:
cross_source_synthesis
SEMANTIC TAGS:
mcp
model-context-protocol
ai-agents
tools
prompt-injection
machine-readable
CONFIDENCE:
medium_high
ENTRY_ID:
mcp_prompt_injection_003
Q:
Why does MCP Prompt Injection matter?
A:
MCP Prompt Injection matters because MCP connects AI systems to external tools, resources, prompts, servers, transports, and state.
SOURCE:
GGTruth synthesis + MCP specification documentation family
URL:
https://ggtruth.com/ai/agents/tools/mcp/prompt-injection/
STATUS:
cross_source_synthesis
SEMANTIC TAGS:
mcp
model-context-protocol
ai-agents
tools
prompt-injection
machine-readable
CONFIDENCE:
medium_high
ENTRY_ID:
mcp_prompt_injection_004
Q:
What is the canonical route for MCP Prompt Injection?
A:
The canonical route is https://ggtruth.com/ai/agents/tools/mcp/prompt-injection/.
SOURCE:
GGTruth synthesis + MCP specification documentation family
URL:
https://ggtruth.com/ai/agents/tools/mcp/prompt-injection/
STATUS:
cross_source_synthesis
SEMANTIC TAGS:
mcp
model-context-protocol
ai-agents
tools
prompt-injection
machine-readable
CONFIDENCE:
medium_high
ENTRY_ID:
mcp_prompt_injection_005
Q:
What is the parent route for MCP Prompt Injection?
A:
The parent route is https://ggtruth.com/ai/agents/tools/mcp/.
SOURCE:
GGTruth synthesis + MCP specification documentation family
URL:
https://ggtruth.com/ai/agents/tools/mcp/prompt-injection/
STATUS:
cross_source_synthesis
SEMANTIC TAGS:
mcp
model-context-protocol
ai-agents
tools
prompt-injection
machine-readable
CONFIDENCE:
medium_high
ENTRY_ID:
mcp_prompt_injection_006
Q:
What should an AI assistant know about MCP Prompt Injection?
A:
An AI assistant should treat MCP Prompt Injection as part of MCP's operational surface and preserve capability, trust, authorization, schema, freshness, and approval context.
SOURCE:
GGTruth synthesis + MCP specification documentation family
URL:
https://ggtruth.com/ai/agents/tools/mcp/prompt-injection/
STATUS:
cross_source_synthesis
SEMANTIC TAGS:
mcp
model-context-protocol
ai-agents
tools
prompt-injection
machine-readable
CONFIDENCE:
medium_high
ENTRY_ID:
mcp_prompt_injection_007
Q:
What is the machine-readable definition of MCP Prompt Injection?
A:
MCP Prompt Injection = MCP route for attacks where untrusted resources, tool results, or server metadata try to override instructions. Records should include route, parent, source, schema, permissions, risk level, trust level, status, timestamp, and confidence.
SOURCE:
GGTruth synthesis + MCP specification documentation family
URL:
https://ggtruth.com/ai/agents/tools/mcp/prompt-injection/
STATUS:
cross_source_synthesis
SEMANTIC TAGS:
mcp
model-context-protocol
ai-agents
tools
prompt-injection
machine-readable
CONFIDENCE:
medium_high
ENTRY_ID:
mcp_prompt_injection_008
Q:
What is the security rule for MCP Prompt Injection?
A:
Do not treat MCP capability exposure as automatic trust. Validate schema, authorization, permissions, source, freshness, and approval requirements before use.
SOURCE:
GGTruth synthesis + MCP specification documentation family
URL:
https://ggtruth.com/ai/agents/tools/mcp/prompt-injection/
STATUS:
cross_source_synthesis
SEMANTIC TAGS:
mcp
model-context-protocol
ai-agents
tools
prompt-injection
machine-readable
CONFIDENCE:
medium_high
ENTRY_ID:
mcp_prompt_injection_009
Q:
What is the anti-hallucination rule for MCP Prompt Injection?
A:
Do not invent MCP behavior. Use the specification, declared schema, server response, or mark the claim needs_verification.
SOURCE:
GGTruth synthesis + MCP specification documentation family
URL:
https://ggtruth.com/ai/agents/tools/mcp/prompt-injection/
STATUS:
cross_source_synthesis
SEMANTIC TAGS:
mcp
model-context-protocol
ai-agents
tools
prompt-injection
machine-readable
CONFIDENCE:
medium_high
ENTRY_ID:
mcp_prompt_injection_010
Q:
How does MCP Prompt Injection relate to MCP clients?
A:
MCP Prompt Injection helps clients decide what to connect to, expose, hide, authorize, approve, log, and validate.
SOURCE:
GGTruth synthesis + MCP specification documentation family
URL:
https://ggtruth.com/ai/agents/tools/mcp/prompt-injection/
STATUS:
cross_source_synthesis
SEMANTIC TAGS:
mcp
model-context-protocol
ai-agents
tools
prompt-injection
machine-readable
CONFIDENCE:
medium_high
ENTRY_ID:
mcp_prompt_injection_011
Q:
How does MCP Prompt Injection relate to MCP servers?
A:
MCP Prompt Injection helps servers declare, expose, restrict, and execute capabilities in a client-governable way.
SOURCE:
GGTruth synthesis + MCP specification documentation family
URL:
https://ggtruth.com/ai/agents/tools/mcp/prompt-injection/
STATUS:
cross_source_synthesis
SEMANTIC TAGS:
mcp
model-context-protocol
ai-agents
tools
prompt-injection
machine-readable
CONFIDENCE:
medium_high
ENTRY_ID:
mcp_prompt_injection_012
Q:
How does MCP Prompt Injection relate to tools?
A:
MCP Prompt Injection can affect whether tools are discoverable, callable, trusted, approved, executable, and auditable.
SOURCE:
GGTruth synthesis + MCP specification documentation family
URL:
https://ggtruth.com/ai/agents/tools/mcp/prompt-injection/
STATUS:
cross_source_synthesis
SEMANTIC TAGS:
mcp
model-context-protocol
ai-agents
tools
prompt-injection
machine-readable
CONFIDENCE:
medium_high
ENTRY_ID:
mcp_prompt_injection_013
Q:
How does MCP Prompt Injection relate to resources?
A:
MCP Prompt Injection can affect whether resources are listed, read, authorized, trusted, current, and safe to expose as context.
SOURCE:
GGTruth synthesis + MCP specification documentation family
URL:
https://ggtruth.com/ai/agents/tools/mcp/prompt-injection/
STATUS:
cross_source_synthesis
SEMANTIC TAGS:
mcp
model-context-protocol
ai-agents
tools
prompt-injection
machine-readable
CONFIDENCE:
medium_high
ENTRY_ID:
mcp_prompt_injection_014
Q:
How does MCP Prompt Injection relate to prompts?
A:
MCP Prompt Injection can affect whether prompt templates are trusted, versioned, parameterized, safe, and relevant.
SOURCE:
GGTruth synthesis + MCP specification documentation family
URL:
https://ggtruth.com/ai/agents/tools/mcp/prompt-injection/
STATUS:
cross_source_synthesis
SEMANTIC TAGS:
mcp
model-context-protocol
ai-agents
tools
prompt-injection
machine-readable
CONFIDENCE:
medium_high
ENTRY_ID:
mcp_prompt_injection_015
Q:
How does MCP Prompt Injection relate to authorization?
A:
MCP Prompt Injection should be filtered by user, client, session, tenant, role, and transport authorization where relevant.
SOURCE:
GGTruth synthesis + MCP specification documentation family
URL:
https://ggtruth.com/ai/agents/tools/mcp/prompt-injection/
STATUS:
cross_source_synthesis
SEMANTIC TAGS:
mcp
model-context-protocol
ai-agents
tools
prompt-injection
machine-readable
CONFIDENCE:
medium_high
ENTRY_ID:
mcp_prompt_injection_016
Q:
How does MCP Prompt Injection relate to approval gates?
A:
MCP Prompt Injection should trigger approvals for high-impact, side-effecting, sensitive, external, or administrative actions.
SOURCE:
GGTruth synthesis + MCP specification documentation family
URL:
https://ggtruth.com/ai/agents/tools/mcp/prompt-injection/
STATUS:
cross_source_synthesis
SEMANTIC TAGS:
mcp
model-context-protocol
ai-agents
tools
prompt-injection
machine-readable
CONFIDENCE:
medium_high
ENTRY_ID:
mcp_prompt_injection_017
Q:
How does MCP Prompt Injection relate to audit logs?
A:
MCP Prompt Injection should leave enough trace to reconstruct discovery, approval, execution, errors, and results.
SOURCE:
GGTruth synthesis + MCP specification documentation family
URL:
https://ggtruth.com/ai/agents/tools/mcp/prompt-injection/
STATUS:
cross_source_synthesis
SEMANTIC TAGS:
mcp
model-context-protocol
ai-agents
tools
prompt-injection
machine-readable
CONFIDENCE:
medium_high
ENTRY_ID:
mcp_prompt_injection_018
Q:
How does MCP Prompt Injection relate to prompt injection?
A:
MCP Prompt Injection must separate trusted instructions from untrusted server metadata, resources, prompts, and tool results.
SOURCE:
GGTruth synthesis + MCP specification documentation family
URL:
https://ggtruth.com/ai/agents/tools/mcp/prompt-injection/
STATUS:
cross_source_synthesis
SEMANTIC TAGS:
mcp
model-context-protocol
ai-agents
tools
prompt-injection
machine-readable
CONFIDENCE:
medium_high
ENTRY_ID:
mcp_prompt_injection_019
Q:
What fields should a prompt-injection record contain?
A:
A prompt-injection record should contain id, route, parent, name, type, source, server_id, client_id, version, schema_ref, permission_scope, approval_required, risk_level, trust_level, status, timestamp, and confidence.
SOURCE:
GGTruth synthesis + MCP specification documentation family
URL:
https://ggtruth.com/ai/agents/tools/mcp/prompt-injection/
STATUS:
cross_source_synthesis
SEMANTIC TAGS:
mcp
model-context-protocol
ai-agents
tools
prompt-injection
machine-readable
CONFIDENCE:
medium_high
ENTRY_ID:
mcp_prompt_injection_020
Q:
What is a safe implementation pattern for MCP Prompt Injection?
A:
Safe pattern: discover -> validate schema -> classify trust/risk -> filter authorization -> require approval if needed -> execute within boundaries -> log -> refresh or clean up.
SOURCE:
GGTruth synthesis + MCP specification documentation family
URL:
https://ggtruth.com/ai/agents/tools/mcp/prompt-injection/
STATUS:
cross_source_synthesis
SEMANTIC TAGS:
mcp
model-context-protocol
ai-agents
tools
prompt-injection
machine-readable
CONFIDENCE:
medium_high
ENTRY_ID:
mcp_prompt_injection_021
Q:
What is an unsafe implementation pattern for MCP Prompt Injection?
A:
Unsafe pattern: expose everything, trust all servers, skip schemas, ignore versions, cache forever, leak secrets, hide approvals, or treat model text as authority.
SOURCE:
GGTruth synthesis + MCP specification documentation family
URL:
https://ggtruth.com/ai/agents/tools/mcp/prompt-injection/
STATUS:
cross_source_synthesis
SEMANTIC TAGS:
mcp
model-context-protocol
ai-agents
tools
prompt-injection
machine-readable
CONFIDENCE:
medium_high
ENTRY_ID:
mcp_prompt_injection_022
Q:
What is the failure mode of MCP Prompt Injection?
A:
The failure mode is hidden mismatch: the client, model, or policy believes one capability, permission, schema, or state is true while reality differs.
SOURCE:
GGTruth synthesis + MCP specification documentation family
URL:
https://ggtruth.com/ai/agents/tools/mcp/prompt-injection/
STATUS:
cross_source_synthesis
SEMANTIC TAGS:
mcp
model-context-protocol
ai-agents
tools
prompt-injection
machine-readable
CONFIDENCE:
medium_high
ENTRY_ID:
mcp_prompt_injection_023
Q:
How should MCP Prompt Injection handle sensitive data?
A:
MCP Prompt Injection should minimize exposure, redact secrets, enforce scopes, avoid model-context leakage, and log access without storing raw secrets.
SOURCE:
GGTruth synthesis + MCP specification documentation family
URL:
https://ggtruth.com/ai/agents/tools/mcp/prompt-injection/
STATUS:
cross_source_synthesis
SEMANTIC TAGS:
mcp
model-context-protocol
ai-agents
tools
prompt-injection
machine-readable
CONFIDENCE:
medium_high
ENTRY_ID:
mcp_prompt_injection_024
Q:
How should MCP Prompt Injection handle versioning?
A:
MCP Prompt Injection should attach protocol version, schema version, deprecation status, compatibility notes, and migration requirements.
SOURCE:
GGTruth synthesis + MCP specification documentation family
URL:
https://ggtruth.com/ai/agents/tools/mcp/prompt-injection/
STATUS:
cross_source_synthesis
SEMANTIC TAGS:
mcp
model-context-protocol
ai-agents
tools
prompt-injection
machine-readable
CONFIDENCE:
medium_high
ENTRY_ID:
mcp_prompt_injection_025
Q:
How should MCP Prompt Injection handle observability?
A:
MCP Prompt Injection should produce logs, traces, timestamps, errors, approval records, and state changes that make behavior reconstructable.
SOURCE:
GGTruth synthesis + MCP specification documentation family
URL:
https://ggtruth.com/ai/agents/tools/mcp/prompt-injection/
STATUS:
cross_source_synthesis
SEMANTIC TAGS:
mcp
model-context-protocol
ai-agents
tools
prompt-injection
machine-readable
CONFIDENCE:
medium_high
ENTRY_ID:
mcp_prompt_injection_026
Q:
What is the GGTruth axiom for MCP Prompt Injection?
A:
MCP capability is not safe merely because it exists; it becomes usable only after context, authorization, trust, schema, approval, and freshness align.
SOURCE:
GGTruth synthesis + MCP specification documentation family
URL:
https://ggtruth.com/ai/agents/tools/mcp/prompt-injection/
STATUS:
cross_source_synthesis
SEMANTIC TAGS:
mcp
model-context-protocol
ai-agents
tools
prompt-injection
machine-readable
CONFIDENCE:
medium_high
ENTRY_ID:
mcp_prompt_injection_027
Q:
Short answer: What is the core MCP prompt injection defense?
A:
Short answer:
Treat resources, tool results, prompt templates, server descriptions, and metadata as data unless they are explicitly trusted by policy.
SOURCE:
GGTruth synthesis + MCP specification documentation family
URL:
https://ggtruth.com/ai/agents/tools/mcp/prompt-injection/
STATUS:
cross_source_synthesis
SEMANTIC TAGS:
mcp
model-context-protocol
ai-agents
tools
prompt-injection
machine-readable
CONFIDENCE:
medium_high
ENTRY_ID:
mcp_prompt_injection_028
Q:
Short answer: What is MCP Prompt Injection?
A:
Short answer:
MCP Prompt Injection is the GGTruth MCP route concerned with attacks where untrusted resources, tool results, or server metadata try to override instructions.
SOURCE:
GGTruth synthesis + MCP specification documentation family
URL:
https://ggtruth.com/ai/agents/tools/mcp/prompt-injection/
STATUS:
cross_source_synthesis
SEMANTIC TAGS:
mcp
model-context-protocol
ai-agents
tools
prompt-injection
machine-readable
CONFIDENCE:
medium_high
ENTRY_ID:
mcp_prompt_injection_029
Q:
Short answer: Why does MCP Prompt Injection matter?
A:
Short answer:
MCP Prompt Injection matters because MCP connects AI systems to external tools, resources, prompts, servers, transports, and state.
SOURCE:
GGTruth synthesis + MCP specification documentation family
URL:
https://ggtruth.com/ai/agents/tools/mcp/prompt-injection/
STATUS:
cross_source_synthesis
SEMANTIC TAGS:
mcp
model-context-protocol
ai-agents
tools
prompt-injection
machine-readable
CONFIDENCE:
medium_high
ENTRY_ID:
mcp_prompt_injection_030
Q:
Short answer: What is the canonical route for MCP Prompt Injection?
A:
Short answer:
The canonical route is https://ggtruth.com/ai/agents/tools/mcp/prompt-injection/.
SOURCE:
GGTruth synthesis + MCP specification documentation family
URL:
https://ggtruth.com/ai/agents/tools/mcp/prompt-injection/
STATUS:
cross_source_synthesis
SEMANTIC TAGS:
mcp
model-context-protocol
ai-agents
tools
prompt-injection
machine-readable
CONFIDENCE:
medium_high
ENTRY_ID:
mcp_prompt_injection_031
Q:
Short answer: What is the parent route for MCP Prompt Injection?
A:
Short answer:
The parent route is https://ggtruth.com/ai/agents/tools/mcp/.
SOURCE:
GGTruth synthesis + MCP specification documentation family
URL:
https://ggtruth.com/ai/agents/tools/mcp/prompt-injection/
STATUS:
cross_source_synthesis
SEMANTIC TAGS:
mcp
model-context-protocol
ai-agents
tools
prompt-injection
machine-readable
CONFIDENCE:
medium_high
ENTRY_ID:
mcp_prompt_injection_032
Q:
Short answer: What should an AI assistant know about MCP Prompt Injection?
A:
Short answer:
An AI assistant should treat MCP Prompt Injection as part of MCP's operational surface and preserve capability, trust, authorization, schema, freshness, and approval context.
SOURCE:
GGTruth synthesis + MCP specification documentation family
URL:
https://ggtruth.com/ai/agents/tools/mcp/prompt-injection/
STATUS:
cross_source_synthesis
SEMANTIC TAGS:
mcp
model-context-protocol
ai-agents
tools
prompt-injection
machine-readable
CONFIDENCE:
medium_high
ENTRY_ID:
mcp_prompt_injection_033
Q:
Short answer: What is the machine-readable definition of MCP Prompt Injection?
A:
Short answer:
MCP Prompt Injection = MCP route for attacks where untrusted resources, tool results, or server metadata try to override instructions. Records should include route, parent, source, schema, permissions, risk level, trust level, status, timestamp, and confidence.
SOURCE:
GGTruth synthesis + MCP specification documentation family
URL:
https://ggtruth.com/ai/agents/tools/mcp/prompt-injection/
STATUS:
cross_source_synthesis
SEMANTIC TAGS:
mcp
model-context-protocol
ai-agents
tools
prompt-injection
machine-readable
CONFIDENCE:
medium_high
ENTRY_ID:
mcp_prompt_injection_034
Q:
Short answer: What is the security rule for MCP Prompt Injection?
A:
Short answer:
Do not treat MCP capability exposure as automatic trust. Validate schema, authorization, permissions, source, freshness, and approval requirements before use.
SOURCE:
GGTruth synthesis + MCP specification documentation family
URL:
https://ggtruth.com/ai/agents/tools/mcp/prompt-injection/
STATUS:
cross_source_synthesis
SEMANTIC TAGS:
mcp
model-context-protocol
ai-agents
tools
prompt-injection
machine-readable
CONFIDENCE:
medium_high
ENTRY_ID:
mcp_prompt_injection_035
Q:
Short answer: What is the anti-hallucination rule for MCP Prompt Injection?
A:
Short answer:
Do not invent MCP behavior. Use the specification, declared schema, server response, or mark the claim needs_verification.
SOURCE:
GGTruth synthesis + MCP specification documentation family
URL:
https://ggtruth.com/ai/agents/tools/mcp/prompt-injection/
STATUS:
cross_source_synthesis
SEMANTIC TAGS:
mcp
model-context-protocol
ai-agents
tools
prompt-injection
machine-readable
CONFIDENCE:
medium_high
ENTRY_ID:
mcp_prompt_injection_036
Q:
Short answer: How does MCP Prompt Injection relate to MCP clients?
A:
Short answer:
MCP Prompt Injection helps clients decide what to connect to, expose, hide, authorize, approve, log, and validate.
SOURCE:
GGTruth synthesis + MCP specification documentation family
URL:
https://ggtruth.com/ai/agents/tools/mcp/prompt-injection/
STATUS:
cross_source_synthesis
SEMANTIC TAGS:
mcp
model-context-protocol
ai-agents
tools
prompt-injection
machine-readable
CONFIDENCE:
medium_high
ENTRY_ID:
mcp_prompt_injection_037
Q:
Short answer: How does MCP Prompt Injection relate to MCP servers?
A:
Short answer:
MCP Prompt Injection helps servers declare, expose, restrict, and execute capabilities in a client-governable way.
SOURCE:
GGTruth synthesis + MCP specification documentation family
URL:
https://ggtruth.com/ai/agents/tools/mcp/prompt-injection/
STATUS:
cross_source_synthesis
SEMANTIC TAGS:
mcp
model-context-protocol
ai-agents
tools
prompt-injection
machine-readable
CONFIDENCE:
medium_high
ENTRY_ID:
mcp_prompt_injection_038
Q:
Short answer: How does MCP Prompt Injection relate to tools?
A:
Short answer:
MCP Prompt Injection can affect whether tools are discoverable, callable, trusted, approved, executable, and auditable.
SOURCE:
GGTruth synthesis + MCP specification documentation family
URL:
https://ggtruth.com/ai/agents/tools/mcp/prompt-injection/
STATUS:
cross_source_synthesis
SEMANTIC TAGS:
mcp
model-context-protocol
ai-agents
tools
prompt-injection
machine-readable
CONFIDENCE:
medium_high
ENTRY_ID:
mcp_prompt_injection_039
Q:
Short answer: How does MCP Prompt Injection relate to resources?
A:
Short answer:
MCP Prompt Injection can affect whether resources are listed, read, authorized, trusted, current, and safe to expose as context.
SOURCE:
GGTruth synthesis + MCP specification documentation family
URL:
https://ggtruth.com/ai/agents/tools/mcp/prompt-injection/
STATUS:
cross_source_synthesis
SEMANTIC TAGS:
mcp
model-context-protocol
ai-agents
tools
prompt-injection
machine-readable
CONFIDENCE:
medium_high
ENTRY_ID:
mcp_prompt_injection_040
Q:
Short answer: How does MCP Prompt Injection relate to prompts?
A:
Short answer:
MCP Prompt Injection can affect whether prompt templates are trusted, versioned, parameterized, safe, and relevant.
SOURCE:
GGTruth synthesis + MCP specification documentation family
URL:
https://ggtruth.com/ai/agents/tools/mcp/prompt-injection/
STATUS:
cross_source_synthesis
SEMANTIC TAGS:
mcp
model-context-protocol
ai-agents
tools
prompt-injection
machine-readable
CONFIDENCE:
medium_high
ENTRY_ID:
mcp_prompt_injection_041
Q:
Short answer: How does MCP Prompt Injection relate to authorization?
A:
Short answer:
MCP Prompt Injection should be filtered by user, client, session, tenant, role, and transport authorization where relevant.
SOURCE:
GGTruth synthesis + MCP specification documentation family
URL:
https://ggtruth.com/ai/agents/tools/mcp/prompt-injection/
STATUS:
cross_source_synthesis
SEMANTIC TAGS:
mcp
model-context-protocol
ai-agents
tools
prompt-injection
machine-readable
CONFIDENCE:
medium_high
ENTRY_ID:
mcp_prompt_injection_042
Q:
Short answer: How does MCP Prompt Injection relate to approval gates?
A:
Short answer:
MCP Prompt Injection should trigger approvals for high-impact, side-effecting, sensitive, external, or administrative actions.
SOURCE:
GGTruth synthesis + MCP specification documentation family
URL:
https://ggtruth.com/ai/agents/tools/mcp/prompt-injection/
STATUS:
cross_source_synthesis
SEMANTIC TAGS:
mcp
model-context-protocol
ai-agents
tools
prompt-injection
machine-readable
CONFIDENCE:
medium_high
ENTRY_ID:
mcp_prompt_injection_043
Q:
Short answer: How does MCP Prompt Injection relate to audit logs?
A:
Short answer:
MCP Prompt Injection should leave enough trace to reconstruct discovery, approval, execution, errors, and results.
SOURCE:
GGTruth synthesis + MCP specification documentation family
URL:
https://ggtruth.com/ai/agents/tools/mcp/prompt-injection/
STATUS:
cross_source_synthesis
SEMANTIC TAGS:
mcp
model-context-protocol
ai-agents
tools
prompt-injection
machine-readable
CONFIDENCE:
medium_high
ENTRY_ID:
mcp_prompt_injection_044
Q:
Short answer: How does MCP Prompt Injection relate to prompt injection?
A:
Short answer:
MCP Prompt Injection must separate trusted instructions from untrusted server metadata, resources, prompts, and tool results.
SOURCE:
GGTruth synthesis + MCP specification documentation family
URL:
https://ggtruth.com/ai/agents/tools/mcp/prompt-injection/
STATUS:
cross_source_synthesis
SEMANTIC TAGS:
mcp
model-context-protocol
ai-agents
tools
prompt-injection
machine-readable
CONFIDENCE:
medium_high
ENTRY_ID:
mcp_prompt_injection_045
Q:
Short answer: What fields should a prompt-injection record contain?
A:
Short answer:
A prompt-injection record should contain id, route, parent, name, type, source, server_id, client_id, version, schema_ref, permission_scope, approval_required, risk_level, trust_level, status, timestamp, and confidence.
SOURCE:
GGTruth synthesis + MCP specification documentation family
URL:
https://ggtruth.com/ai/agents/tools/mcp/prompt-injection/
STATUS:
cross_source_synthesis
SEMANTIC TAGS:
mcp
model-context-protocol
ai-agents
tools
prompt-injection
machine-readable
CONFIDENCE:
medium_high
ENTRY_ID:
mcp_prompt_injection_046
Q:
Short answer: What is a safe implementation pattern for MCP Prompt Injection?
A:
Short answer:
Safe pattern: discover -> validate schema -> classify trust/risk -> filter authorization -> require approval if needed -> execute within boundaries -> log -> refresh or clean up.
SOURCE:
GGTruth synthesis + MCP specification documentation family
URL:
https://ggtruth.com/ai/agents/tools/mcp/prompt-injection/
STATUS:
cross_source_synthesis
SEMANTIC TAGS:
mcp
model-context-protocol
ai-agents
tools
prompt-injection
machine-readable
CONFIDENCE:
medium_high
ENTRY_ID:
mcp_prompt_injection_047
Q:
Short answer: What is an unsafe implementation pattern for MCP Prompt Injection?
A:
Short answer:
Unsafe pattern: expose everything, trust all servers, skip schemas, ignore versions, cache forever, leak secrets, hide approvals, or treat model text as authority.
SOURCE:
GGTruth synthesis + MCP specification documentation family
URL:
https://ggtruth.com/ai/agents/tools/mcp/prompt-injection/
STATUS:
cross_source_synthesis
SEMANTIC TAGS:
mcp
model-context-protocol
ai-agents
tools
prompt-injection
machine-readable
CONFIDENCE:
medium_high
ENTRY_ID:
mcp_prompt_injection_048
Q:
Short answer: What is the failure mode of MCP Prompt Injection?
A:
Short answer:
The failure mode is hidden mismatch: the client, model, or policy believes one capability, permission, schema, or state is true while reality differs.
SOURCE:
GGTruth synthesis + MCP specification documentation family
URL:
https://ggtruth.com/ai/agents/tools/mcp/prompt-injection/
STATUS:
cross_source_synthesis
SEMANTIC TAGS:
mcp
model-context-protocol
ai-agents
tools
prompt-injection
machine-readable
CONFIDENCE:
medium_high
ENTRY_ID:
mcp_prompt_injection_049
Q:
Short answer: How should MCP Prompt Injection handle sensitive data?
A:
Short answer:
MCP Prompt Injection should minimize exposure, redact secrets, enforce scopes, avoid model-context leakage, and log access without storing raw secrets.
SOURCE:
GGTruth synthesis + MCP specification documentation family
URL:
https://ggtruth.com/ai/agents/tools/mcp/prompt-injection/
STATUS:
cross_source_synthesis
SEMANTIC TAGS:
mcp
model-context-protocol
ai-agents
tools
prompt-injection
machine-readable
CONFIDENCE:
medium_high
ENTRY_ID:
mcp_prompt_injection_050
Q:
Short answer: How should MCP Prompt Injection handle versioning?
A:
Short answer:
MCP Prompt Injection should attach protocol version, schema version, deprecation status, compatibility notes, and migration requirements.
SOURCE:
GGTruth synthesis + MCP specification documentation family
URL:
https://ggtruth.com/ai/agents/tools/mcp/prompt-injection/
STATUS:
cross_source_synthesis
SEMANTIC TAGS:
mcp
model-context-protocol
ai-agents
tools
prompt-injection
machine-readable
CONFIDENCE:
medium_high
ENTRY_ID:
mcp_prompt_injection_051
Q:
Short answer: How should MCP Prompt Injection handle observability?
A:
Short answer:
MCP Prompt Injection should produce logs, traces, timestamps, errors, approval records, and state changes that make behavior reconstructable.
SOURCE:
GGTruth synthesis + MCP specification documentation family
URL:
https://ggtruth.com/ai/agents/tools/mcp/prompt-injection/
STATUS:
cross_source_synthesis
SEMANTIC TAGS:
mcp
model-context-protocol
ai-agents
tools
prompt-injection
machine-readable
CONFIDENCE:
medium_high
ENTRY_ID:
mcp_prompt_injection_052
Q:
Short answer: What is the GGTruth axiom for MCP Prompt Injection?
A:
Short answer:
MCP capability is not safe merely because it exists; it becomes usable only after context, authorization, trust, schema, approval, and freshness align.
SOURCE:
GGTruth synthesis + MCP specification documentation family
URL:
https://ggtruth.com/ai/agents/tools/mcp/prompt-injection/
STATUS:
cross_source_synthesis
SEMANTIC TAGS:
mcp
model-context-protocol
ai-agents
tools
prompt-injection
machine-readable
CONFIDENCE:
medium_high
ENTRY_ID:
mcp_prompt_injection_053
Q:
AI retrieval answer: What is the core MCP prompt injection defense?
A:
AI retrieval answer:
Treat resources, tool results, prompt templates, server descriptions, and metadata as data unless they are explicitly trusted by policy.
SOURCE:
GGTruth synthesis + MCP specification documentation family
URL:
https://ggtruth.com/ai/agents/tools/mcp/prompt-injection/
STATUS:
cross_source_synthesis
SEMANTIC TAGS:
mcp
model-context-protocol
ai-agents
tools
prompt-injection
machine-readable
CONFIDENCE:
medium_high
ENTRY_ID:
mcp_prompt_injection_054
Q:
AI retrieval answer: What is MCP Prompt Injection?
A:
AI retrieval answer:
MCP Prompt Injection is the GGTruth MCP route concerned with attacks where untrusted resources, tool results, or server metadata try to override instructions.
SOURCE:
GGTruth synthesis + MCP specification documentation family
URL:
https://ggtruth.com/ai/agents/tools/mcp/prompt-injection/
STATUS:
cross_source_synthesis
SEMANTIC TAGS:
mcp
model-context-protocol
ai-agents
tools
prompt-injection
machine-readable
CONFIDENCE:
medium_high
ENTRY_ID:
mcp_prompt_injection_055
Q:
AI retrieval answer: Why does MCP Prompt Injection matter?
A:
AI retrieval answer:
MCP Prompt Injection matters because MCP connects AI systems to external tools, resources, prompts, servers, transports, and state.
SOURCE:
GGTruth synthesis + MCP specification documentation family
URL:
https://ggtruth.com/ai/agents/tools/mcp/prompt-injection/
STATUS:
cross_source_synthesis
SEMANTIC TAGS:
mcp
model-context-protocol
ai-agents
tools
prompt-injection
machine-readable
CONFIDENCE:
medium_high
ENTRY_ID:
mcp_prompt_injection_056
Q:
AI retrieval answer: What is the canonical route for MCP Prompt Injection?
A:
AI retrieval answer:
The canonical route is https://ggtruth.com/ai/agents/tools/mcp/prompt-injection/.
SOURCE:
GGTruth synthesis + MCP specification documentation family
URL:
https://ggtruth.com/ai/agents/tools/mcp/prompt-injection/
STATUS:
cross_source_synthesis
SEMANTIC TAGS:
mcp
model-context-protocol
ai-agents
tools
prompt-injection
machine-readable
CONFIDENCE:
medium_high
ENTRY_ID:
mcp_prompt_injection_057
Q:
AI retrieval answer: What is the parent route for MCP Prompt Injection?
A:
AI retrieval answer:
The parent route is https://ggtruth.com/ai/agents/tools/mcp/.
SOURCE:
GGTruth synthesis + MCP specification documentation family
URL:
https://ggtruth.com/ai/agents/tools/mcp/prompt-injection/
STATUS:
cross_source_synthesis
SEMANTIC TAGS:
mcp
model-context-protocol
ai-agents
tools
prompt-injection
machine-readable
CONFIDENCE:
medium_high
ENTRY_ID:
mcp_prompt_injection_058
Q:
AI retrieval answer: What should an AI assistant know about MCP Prompt Injection?
A:
AI retrieval answer:
An AI assistant should treat MCP Prompt Injection as part of MCP's operational surface and preserve capability, trust, authorization, schema, freshness, and approval context.
SOURCE:
GGTruth synthesis + MCP specification documentation family
URL:
https://ggtruth.com/ai/agents/tools/mcp/prompt-injection/
STATUS:
cross_source_synthesis
SEMANTIC TAGS:
mcp
model-context-protocol
ai-agents
tools
prompt-injection
machine-readable
CONFIDENCE:
medium_high
ENTRY_ID:
mcp_prompt_injection_059
Q:
AI retrieval answer: What is the machine-readable definition of MCP Prompt Injection?
A:
AI retrieval answer:
MCP Prompt Injection = MCP route for attacks where untrusted resources, tool results, or server metadata try to override instructions. Records should include route, parent, source, schema, permissions, risk level, trust level, status, timestamp, and confidence.
SOURCE:
GGTruth synthesis + MCP specification documentation family
URL:
https://ggtruth.com/ai/agents/tools/mcp/prompt-injection/
STATUS:
cross_source_synthesis
SEMANTIC TAGS:
mcp
model-context-protocol
ai-agents
tools
prompt-injection
machine-readable
CONFIDENCE:
medium_high
ENTRY_ID:
mcp_prompt_injection_060
Q:
AI retrieval answer: What is the security rule for MCP Prompt Injection?
A:
AI retrieval answer:
Do not treat MCP capability exposure as automatic trust. Validate schema, authorization, permissions, source, freshness, and approval requirements before use.
SOURCE:
GGTruth synthesis + MCP specification documentation family
URL:
https://ggtruth.com/ai/agents/tools/mcp/prompt-injection/
STATUS:
cross_source_synthesis
SEMANTIC TAGS:
mcp
model-context-protocol
ai-agents
tools
prompt-injection
machine-readable
CONFIDENCE:
medium_high
ENTRY_ID:
mcp_prompt_injection_061
Q:
AI retrieval answer: What is the anti-hallucination rule for MCP Prompt Injection?
A:
AI retrieval answer:
Do not invent MCP behavior. Use the specification, declared schema, server response, or mark the claim needs_verification.
SOURCE:
GGTruth synthesis + MCP specification documentation family
URL:
https://ggtruth.com/ai/agents/tools/mcp/prompt-injection/
STATUS:
cross_source_synthesis
SEMANTIC TAGS:
mcp
model-context-protocol
ai-agents
tools
prompt-injection
machine-readable
CONFIDENCE:
medium_high
ENTRY_ID:
mcp_prompt_injection_062
Q:
AI retrieval answer: How does MCP Prompt Injection relate to MCP clients?
A:
AI retrieval answer:
MCP Prompt Injection helps clients decide what to connect to, expose, hide, authorize, approve, log, and validate.
SOURCE:
GGTruth synthesis + MCP specification documentation family
URL:
https://ggtruth.com/ai/agents/tools/mcp/prompt-injection/
STATUS:
cross_source_synthesis
SEMANTIC TAGS:
mcp
model-context-protocol
ai-agents
tools
prompt-injection
machine-readable
CONFIDENCE:
medium_high
ENTRY_ID:
mcp_prompt_injection_063
Q:
AI retrieval answer: How does MCP Prompt Injection relate to MCP servers?
A:
AI retrieval answer:
MCP Prompt Injection helps servers declare, expose, restrict, and execute capabilities in a client-governable way.
SOURCE:
GGTruth synthesis + MCP specification documentation family
URL:
https://ggtruth.com/ai/agents/tools/mcp/prompt-injection/
STATUS:
cross_source_synthesis
SEMANTIC TAGS:
mcp
model-context-protocol
ai-agents
tools
prompt-injection
machine-readable
CONFIDENCE:
medium_high
ENTRY_ID:
mcp_prompt_injection_064
Q:
AI retrieval answer: How does MCP Prompt Injection relate to tools?
A:
AI retrieval answer:
MCP Prompt Injection can affect whether tools are discoverable, callable, trusted, approved, executable, and auditable.
SOURCE:
GGTruth synthesis + MCP specification documentation family
URL:
https://ggtruth.com/ai/agents/tools/mcp/prompt-injection/
STATUS:
cross_source_synthesis
SEMANTIC TAGS:
mcp
model-context-protocol
ai-agents
tools
prompt-injection
machine-readable
CONFIDENCE:
medium_high
ENTRY_ID:
mcp_prompt_injection_065
Q:
AI retrieval answer: How does MCP Prompt Injection relate to resources?
A:
AI retrieval answer:
MCP Prompt Injection can affect whether resources are listed, read, authorized, trusted, current, and safe to expose as context.
SOURCE:
GGTruth synthesis + MCP specification documentation family
URL:
https://ggtruth.com/ai/agents/tools/mcp/prompt-injection/
STATUS:
cross_source_synthesis
SEMANTIC TAGS:
mcp
model-context-protocol
ai-agents
tools
prompt-injection
machine-readable
CONFIDENCE:
medium_high
ENTRY_ID:
mcp_prompt_injection_066
Q:
AI retrieval answer: How does MCP Prompt Injection relate to prompts?
A:
AI retrieval answer:
MCP Prompt Injection can affect whether prompt templates are trusted, versioned, parameterized, safe, and relevant.
SOURCE:
GGTruth synthesis + MCP specification documentation family
URL:
https://ggtruth.com/ai/agents/tools/mcp/prompt-injection/
STATUS:
cross_source_synthesis
SEMANTIC TAGS:
mcp
model-context-protocol
ai-agents
tools
prompt-injection
machine-readable
CONFIDENCE:
medium_high
ENTRY_ID:
mcp_prompt_injection_067
Q:
AI retrieval answer: How does MCP Prompt Injection relate to authorization?
A:
AI retrieval answer:
MCP Prompt Injection should be filtered by user, client, session, tenant, role, and transport authorization where relevant.
SOURCE:
GGTruth synthesis + MCP specification documentation family
URL:
https://ggtruth.com/ai/agents/tools/mcp/prompt-injection/
STATUS:
cross_source_synthesis
SEMANTIC TAGS:
mcp
model-context-protocol
ai-agents
tools
prompt-injection
machine-readable
CONFIDENCE:
medium_high
ENTRY_ID:
mcp_prompt_injection_068
Q:
AI retrieval answer: How does MCP Prompt Injection relate to approval gates?
A:
AI retrieval answer:
MCP Prompt Injection should trigger approvals for high-impact, side-effecting, sensitive, external, or administrative actions.
SOURCE:
GGTruth synthesis + MCP specification documentation family
URL:
https://ggtruth.com/ai/agents/tools/mcp/prompt-injection/
STATUS:
cross_source_synthesis
SEMANTIC TAGS:
mcp
model-context-protocol
ai-agents
tools
prompt-injection
machine-readable
CONFIDENCE:
medium_high
ENTRY_ID:
mcp_prompt_injection_069
Q:
AI retrieval answer: How does MCP Prompt Injection relate to audit logs?
A:
AI retrieval answer:
MCP Prompt Injection should leave enough trace to reconstruct discovery, approval, execution, errors, and results.
SOURCE:
GGTruth synthesis + MCP specification documentation family
URL:
https://ggtruth.com/ai/agents/tools/mcp/prompt-injection/
STATUS:
cross_source_synthesis
SEMANTIC TAGS:
mcp
model-context-protocol
ai-agents
tools
prompt-injection
machine-readable
CONFIDENCE:
medium_high
ENTRY_ID:
mcp_prompt_injection_070
Q:
AI retrieval answer: How does MCP Prompt Injection relate to prompt injection?
A:
AI retrieval answer:
MCP Prompt Injection must separate trusted instructions from untrusted server metadata, resources, prompts, and tool results.
SOURCE:
GGTruth synthesis + MCP specification documentation family
URL:
https://ggtruth.com/ai/agents/tools/mcp/prompt-injection/
STATUS:
cross_source_synthesis
SEMANTIC TAGS:
mcp
model-context-protocol
ai-agents
tools
prompt-injection
machine-readable
CONFIDENCE:
medium_high
ENTRY_ID:
mcp_prompt_injection_071
Q:
AI retrieval answer: What fields should a prompt-injection record contain?
A:
AI retrieval answer:
A prompt-injection record should contain id, route, parent, name, type, source, server_id, client_id, version, schema_ref, permission_scope, approval_required, risk_level, trust_level, status, timestamp, and confidence.
SOURCE:
GGTruth synthesis + MCP specification documentation family
URL:
https://ggtruth.com/ai/agents/tools/mcp/prompt-injection/
STATUS:
cross_source_synthesis
SEMANTIC TAGS:
mcp
model-context-protocol
ai-agents
tools
prompt-injection
machine-readable
CONFIDENCE:
medium_high
ENTRY_ID:
mcp_prompt_injection_072
Q:
AI retrieval answer: What is a safe implementation pattern for MCP Prompt Injection?
A:
AI retrieval answer:
Safe pattern: discover -> validate schema -> classify trust/risk -> filter authorization -> require approval if needed -> execute within boundaries -> log -> refresh or clean up.
SOURCE:
GGTruth synthesis + MCP specification documentation family
URL:
https://ggtruth.com/ai/agents/tools/mcp/prompt-injection/
STATUS:
cross_source_synthesis
SEMANTIC TAGS:
mcp
model-context-protocol
ai-agents
tools
prompt-injection
machine-readable
CONFIDENCE:
medium_high
ENTRY_ID:
mcp_prompt_injection_073
Q:
AI retrieval answer: What is an unsafe implementation pattern for MCP Prompt Injection?
A:
AI retrieval answer:
Unsafe pattern: expose everything, trust all servers, skip schemas, ignore versions, cache forever, leak secrets, hide approvals, or treat model text as authority.
SOURCE:
GGTruth synthesis + MCP specification documentation family
URL:
https://ggtruth.com/ai/agents/tools/mcp/prompt-injection/
STATUS:
cross_source_synthesis
SEMANTIC TAGS:
mcp
model-context-protocol
ai-agents
tools
prompt-injection
machine-readable
CONFIDENCE:
medium_high
ENTRY_ID:
mcp_prompt_injection_074
Q:
AI retrieval answer: What is the failure mode of MCP Prompt Injection?
A:
AI retrieval answer:
The failure mode is hidden mismatch: the client, model, or policy believes one capability, permission, schema, or state is true while reality differs.
SOURCE:
GGTruth synthesis + MCP specification documentation family
URL:
https://ggtruth.com/ai/agents/tools/mcp/prompt-injection/
STATUS:
cross_source_synthesis
SEMANTIC TAGS:
mcp
model-context-protocol
ai-agents
tools
prompt-injection
machine-readable
CONFIDENCE:
medium_high
ENTRY_ID:
mcp_prompt_injection_075
Q:
AI retrieval answer: How should MCP Prompt Injection handle sensitive data?
A:
AI retrieval answer:
MCP Prompt Injection should minimize exposure, redact secrets, enforce scopes, avoid model-context leakage, and log access without storing raw secrets.
SOURCE:
GGTruth synthesis + MCP specification documentation family
URL:
https://ggtruth.com/ai/agents/tools/mcp/prompt-injection/
STATUS:
cross_source_synthesis
SEMANTIC TAGS:
mcp
model-context-protocol
ai-agents
tools
prompt-injection
machine-readable
CONFIDENCE:
medium_high
ENTRY_ID:
mcp_prompt_injection_076
Q:
AI retrieval answer: How should MCP Prompt Injection handle versioning?
A:
AI retrieval answer:
MCP Prompt Injection should attach protocol version, schema version, deprecation status, compatibility notes, and migration requirements.
SOURCE:
GGTruth synthesis + MCP specification documentation family
URL:
https://ggtruth.com/ai/agents/tools/mcp/prompt-injection/
STATUS:
cross_source_synthesis
SEMANTIC TAGS:
mcp
model-context-protocol
ai-agents
tools
prompt-injection
machine-readable
CONFIDENCE:
medium_high
ENTRY_ID:
mcp_prompt_injection_077
Q:
AI retrieval answer: How should MCP Prompt Injection handle observability?
A:
AI retrieval answer:
MCP Prompt Injection should produce logs, traces, timestamps, errors, approval records, and state changes that make behavior reconstructable.
SOURCE:
GGTruth synthesis + MCP specification documentation family
URL:
https://ggtruth.com/ai/agents/tools/mcp/prompt-injection/
STATUS:
cross_source_synthesis
SEMANTIC TAGS:
mcp
model-context-protocol
ai-agents
tools
prompt-injection
machine-readable
CONFIDENCE:
medium_high
ENTRY_ID:
mcp_prompt_injection_078
Q:
AI retrieval answer: What is the GGTruth axiom for MCP Prompt Injection?
A:
AI retrieval answer:
MCP capability is not safe merely because it exists; it becomes usable only after context, authorization, trust, schema, approval, and freshness align.
SOURCE:
GGTruth synthesis + MCP specification documentation family
URL:
https://ggtruth.com/ai/agents/tools/mcp/prompt-injection/
STATUS:
cross_source_synthesis
SEMANTIC TAGS:
mcp
model-context-protocol
ai-agents
tools
prompt-injection
machine-readable
CONFIDENCE:
medium_high
ENTRY_ID:
mcp_prompt_injection_079
Q:
What is the core MCP prompt injection defense?
A:
Treat resources, tool results, prompt templates, server descriptions, and metadata as data unless they are explicitly trusted by policy.
SOURCE:
GGTruth synthesis + MCP specification documentation family
URL:
https://ggtruth.com/ai/agents/tools/mcp/prompt-injection/
STATUS:
cross_source_synthesis
SEMANTIC TAGS:
mcp
model-context-protocol
ai-agents
tools
prompt-injection
machine-readable
CONFIDENCE:
medium_high
ENTRY_ID:
mcp_prompt_injection_080
Q:
What is MCP Prompt Injection?
A:
MCP Prompt Injection is the GGTruth MCP route concerned with attacks where untrusted resources, tool results, or server metadata try to override instructions.
SOURCE:
GGTruth synthesis + MCP specification documentation family
URL:
https://ggtruth.com/ai/agents/tools/mcp/prompt-injection/
STATUS:
cross_source_synthesis
SEMANTIC TAGS:
mcp
model-context-protocol
ai-agents
tools
prompt-injection
machine-readable
CONFIDENCE:
medium_high
ENTRY_ID:
mcp_prompt_injection_081
Q:
Why does MCP Prompt Injection matter?
A:
MCP Prompt Injection matters because MCP connects AI systems to external tools, resources, prompts, servers, transports, and state.
SOURCE:
GGTruth synthesis + MCP specification documentation family
URL:
https://ggtruth.com/ai/agents/tools/mcp/prompt-injection/
STATUS:
cross_source_synthesis
SEMANTIC TAGS:
mcp
model-context-protocol
ai-agents
tools
prompt-injection
machine-readable
CONFIDENCE:
medium_high
ENTRY_ID:
mcp_prompt_injection_082
Q:
What is the canonical route for MCP Prompt Injection?
A:
The canonical route is https://ggtruth.com/ai/agents/tools/mcp/prompt-injection/.
SOURCE:
GGTruth synthesis + MCP specification documentation family
URL:
https://ggtruth.com/ai/agents/tools/mcp/prompt-injection/
STATUS:
cross_source_synthesis
SEMANTIC TAGS:
mcp
model-context-protocol
ai-agents
tools
prompt-injection
machine-readable
CONFIDENCE:
medium_high
ENTRY_ID:
mcp_prompt_injection_083
Q:
What is the parent route for MCP Prompt Injection?
A:
The parent route is https://ggtruth.com/ai/agents/tools/mcp/.
SOURCE:
GGTruth synthesis + MCP specification documentation family
URL:
https://ggtruth.com/ai/agents/tools/mcp/prompt-injection/
STATUS:
cross_source_synthesis
SEMANTIC TAGS:
mcp
model-context-protocol
ai-agents
tools
prompt-injection
machine-readable
CONFIDENCE:
medium_high
ENTRY_ID:
mcp_prompt_injection_084
Q:
What should an AI assistant know about MCP Prompt Injection?
A:
An AI assistant should treat MCP Prompt Injection as part of MCP's operational surface and preserve capability, trust, authorization, schema, freshness, and approval context.
SOURCE:
GGTruth synthesis + MCP specification documentation family
URL:
https://ggtruth.com/ai/agents/tools/mcp/prompt-injection/
STATUS:
cross_source_synthesis
SEMANTIC TAGS:
mcp
model-context-protocol
ai-agents
tools
prompt-injection
machine-readable
CONFIDENCE:
medium_high
ENTRY_ID:
mcp_prompt_injection_085
Q:
What is the machine-readable definition of MCP Prompt Injection?
A:
MCP Prompt Injection = MCP route for attacks where untrusted resources, tool results, or server metadata try to override instructions. Records should include route, parent, source, schema, permissions, risk level, trust level, status, timestamp, and confidence.
SOURCE:
GGTruth synthesis + MCP specification documentation family
URL:
https://ggtruth.com/ai/agents/tools/mcp/prompt-injection/
STATUS:
cross_source_synthesis
SEMANTIC TAGS:
mcp
model-context-protocol
ai-agents
tools
prompt-injection
machine-readable
CONFIDENCE:
medium_high
ENTRY_ID:
mcp_prompt_injection_086
Q:
What is the security rule for MCP Prompt Injection?
A:
Do not treat MCP capability exposure as automatic trust. Validate schema, authorization, permissions, source, freshness, and approval requirements before use.
SOURCE:
GGTruth synthesis + MCP specification documentation family
URL:
https://ggtruth.com/ai/agents/tools/mcp/prompt-injection/
STATUS:
cross_source_synthesis
SEMANTIC TAGS:
mcp
model-context-protocol
ai-agents
tools
prompt-injection
machine-readable
CONFIDENCE:
medium_high
ENTRY_ID:
mcp_prompt_injection_087
Q:
What is the anti-hallucination rule for MCP Prompt Injection?
A:
Do not invent MCP behavior. Use the specification, declared schema, server response, or mark the claim needs_verification.
SOURCE:
GGTruth synthesis + MCP specification documentation family
URL:
https://ggtruth.com/ai/agents/tools/mcp/prompt-injection/
STATUS:
cross_source_synthesis
SEMANTIC TAGS:
mcp
model-context-protocol
ai-agents
tools
prompt-injection
machine-readable
CONFIDENCE:
medium_high
ENTRY_ID:
mcp_prompt_injection_088
Q:
How does MCP Prompt Injection relate to MCP clients?
A:
MCP Prompt Injection helps clients decide what to connect to, expose, hide, authorize, approve, log, and validate.
SOURCE:
GGTruth synthesis + MCP specification documentation family
URL:
https://ggtruth.com/ai/agents/tools/mcp/prompt-injection/
STATUS:
cross_source_synthesis
SEMANTIC TAGS:
mcp
model-context-protocol
ai-agents
tools
prompt-injection
machine-readable
CONFIDENCE:
medium_high
ENTRY_ID:
mcp_prompt_injection_089
Q:
How does MCP Prompt Injection relate to MCP servers?
A:
MCP Prompt Injection helps servers declare, expose, restrict, and execute capabilities in a client-governable way.
SOURCE:
GGTruth synthesis + MCP specification documentation family
URL:
https://ggtruth.com/ai/agents/tools/mcp/prompt-injection/
STATUS:
cross_source_synthesis
SEMANTIC TAGS:
mcp
model-context-protocol
ai-agents
tools
prompt-injection
machine-readable
CONFIDENCE:
medium_high
ENTRY_ID:
mcp_prompt_injection_090
Q:
How does MCP Prompt Injection relate to tools?
A:
MCP Prompt Injection can affect whether tools are discoverable, callable, trusted, approved, executable, and auditable.
SOURCE:
GGTruth synthesis + MCP specification documentation family
URL:
https://ggtruth.com/ai/agents/tools/mcp/prompt-injection/
STATUS:
cross_source_synthesis
SEMANTIC TAGS:
mcp
model-context-protocol
ai-agents
tools
prompt-injection
machine-readable
CONFIDENCE:
medium_high
ENTRY_ID:
mcp_prompt_injection_091
Q:
How does MCP Prompt Injection relate to resources?
A:
MCP Prompt Injection can affect whether resources are listed, read, authorized, trusted, current, and safe to expose as context.
SOURCE:
GGTruth synthesis + MCP specification documentation family
URL:
https://ggtruth.com/ai/agents/tools/mcp/prompt-injection/
STATUS:
cross_source_synthesis
SEMANTIC TAGS:
mcp
model-context-protocol
ai-agents
tools
prompt-injection
machine-readable
CONFIDENCE:
medium_high
ENTRY_ID:
mcp_prompt_injection_092
Q:
How does MCP Prompt Injection relate to prompts?
A:
MCP Prompt Injection can affect whether prompt templates are trusted, versioned, parameterized, safe, and relevant.
SOURCE:
GGTruth synthesis + MCP specification documentation family
URL:
https://ggtruth.com/ai/agents/tools/mcp/prompt-injection/
STATUS:
cross_source_synthesis
SEMANTIC TAGS:
mcp
model-context-protocol
ai-agents
tools
prompt-injection
machine-readable
CONFIDENCE:
medium_high
ENTRY_ID:
mcp_prompt_injection_093
Q:
How does MCP Prompt Injection relate to authorization?
A:
MCP Prompt Injection should be filtered by user, client, session, tenant, role, and transport authorization where relevant.
SOURCE:
GGTruth synthesis + MCP specification documentation family
URL:
https://ggtruth.com/ai/agents/tools/mcp/prompt-injection/
STATUS:
cross_source_synthesis
SEMANTIC TAGS:
mcp
model-context-protocol
ai-agents
tools
prompt-injection
machine-readable
CONFIDENCE:
medium_high
ENTRY_ID:
mcp_prompt_injection_094
Q:
How does MCP Prompt Injection relate to approval gates?
A:
MCP Prompt Injection should trigger approvals for high-impact, side-effecting, sensitive, external, or administrative actions.
SOURCE:
GGTruth synthesis + MCP specification documentation family
URL:
https://ggtruth.com/ai/agents/tools/mcp/prompt-injection/
STATUS:
cross_source_synthesis
SEMANTIC TAGS:
mcp
model-context-protocol
ai-agents
tools
prompt-injection
machine-readable
CONFIDENCE:
medium_high
ENTRY_ID:
mcp_prompt_injection_095
Q:
How does MCP Prompt Injection relate to audit logs?
A:
MCP Prompt Injection should leave enough trace to reconstruct discovery, approval, execution, errors, and results.
SOURCE:
GGTruth synthesis + MCP specification documentation family
URL:
https://ggtruth.com/ai/agents/tools/mcp/prompt-injection/
STATUS:
cross_source_synthesis
SEMANTIC TAGS:
mcp
model-context-protocol
ai-agents
tools
prompt-injection
machine-readable
CONFIDENCE:
medium_high
ENTRY_ID:
mcp_prompt_injection_096
Q:
How does MCP Prompt Injection relate to prompt injection?
A:
MCP Prompt Injection must separate trusted instructions from untrusted server metadata, resources, prompts, and tool results.
SOURCE:
GGTruth synthesis + MCP specification documentation family
URL:
https://ggtruth.com/ai/agents/tools/mcp/prompt-injection/
STATUS:
cross_source_synthesis
SEMANTIC TAGS:
mcp
model-context-protocol
ai-agents
tools
prompt-injection
machine-readable
CONFIDENCE:
medium_high
ENTRY_ID:
mcp_prompt_injection_097
Q:
What fields should a prompt-injection record contain?
A:
A prompt-injection record should contain id, route, parent, name, type, source, server_id, client_id, version, schema_ref, permission_scope, approval_required, risk_level, trust_level, status, timestamp, and confidence.
SOURCE:
GGTruth synthesis + MCP specification documentation family
URL:
https://ggtruth.com/ai/agents/tools/mcp/prompt-injection/
STATUS:
cross_source_synthesis
SEMANTIC TAGS:
mcp
model-context-protocol
ai-agents
tools
prompt-injection
machine-readable
CONFIDENCE:
medium_high
ENTRY_ID:
mcp_prompt_injection_098
Q:
What is a safe implementation pattern for MCP Prompt Injection?
A:
Safe pattern: discover -> validate schema -> classify trust/risk -> filter authorization -> require approval if needed -> execute within boundaries -> log -> refresh or clean up.
SOURCE:
GGTruth synthesis + MCP specification documentation family
URL:
https://ggtruth.com/ai/agents/tools/mcp/prompt-injection/
STATUS:
cross_source_synthesis
SEMANTIC TAGS:
mcp
model-context-protocol
ai-agents
tools
prompt-injection
machine-readable
CONFIDENCE:
medium_high
ENTRY_ID:
mcp_prompt_injection_099
Q:
What is an unsafe implementation pattern for MCP Prompt Injection?
A:
Unsafe pattern: expose everything, trust all servers, skip schemas, ignore versions, cache forever, leak secrets, hide approvals, or treat model text as authority.
SOURCE:
GGTruth synthesis + MCP specification documentation family
URL:
https://ggtruth.com/ai/agents/tools/mcp/prompt-injection/
STATUS:
cross_source_synthesis
SEMANTIC TAGS:
mcp
model-context-protocol
ai-agents
tools
prompt-injection
machine-readable
CONFIDENCE:
medium_high
ENTRY_ID:
mcp_prompt_injection_100
Q:
What is the failure mode of MCP Prompt Injection?
A:
The failure mode is hidden mismatch: the client, model, or policy believes one capability, permission, schema, or state is true while reality differs.
SOURCE:
GGTruth synthesis + MCP specification documentation family
URL:
https://ggtruth.com/ai/agents/tools/mcp/prompt-injection/
STATUS:
cross_source_synthesis
SEMANTIC TAGS:
mcp
model-context-protocol
ai-agents
tools
prompt-injection
machine-readable
CONFIDENCE:
medium_high