Skip to content

armgit / AiSOC

Public Imported from GitHub
0

Open-source AI Security Operations Center: alert fusion, LLM-agent triage, MITRE ATT&CK investigation, and a replayable decision ledger for every agent step. Self-hostable, runs with no API keys, MIT licensed. Ships an MCP server for Claude, Cursor and Continue.

Fix: Wire prompt sanitizer into classification agents to prevent prompt injection (#219)

Tanmay Zade25e5f52181a9
diff --git a/services/agents/app/agents/auto_triage_agent.py b/services/agents/app/agents/auto_triage_agent.py
index b4df546e5..8647ef440 100644
--- a/services/agents/app/agents/auto_triage_agent.py
+++ b/services/agents/app/agents/auto_triage_agent.py
@@ -21,6 +21,7 @@ import structlog
from langchain_core.messages import HumanMessage, SystemMessage
from langchain_openai import ChatOpenAI
+from app.investigator.prompt_sanitizer import sanitize_text, wrap_untrusted
from app.llm import safe_ainvoke
from app.models.state import AgentStatus, InvestigationState
from app.prompt_serialization import format_extra_fields_for_llm
@@ -98,9 +99,9 @@ def _build_alert_context(state: InvestigationState) -> str:
"""Serialise the alert into a compact string the LLM can reason over."""
raw = state.raw_alert
parts = [
- f"Alert Summary: {state.alert_summary}",
- f"Severity (vendor): {raw.get('severity', 'unknown')}",
- f"Risk Score (vendor): {raw.get('risk_score', 'N/A')}",
+ f"Alert Summary: {sanitize_text(state.alert_summary)}",
+ f"Severity (vendor): {sanitize_text(str(raw.get('severity', 'unknown')))}",
+ f"Risk Score (vendor): {sanitize_text(str(raw.get('risk_score', 'N/A')))}",
]
ioc_fields = {
@@ -111,7 +112,7 @@ def _build_alert_context(state: InvestigationState) -> str:
"url": "URL",
"hostname": "Hostname",
}
- present_iocs = {label: raw[key] for key, label in ioc_fields.items() if raw.get(key)}
+ present_iocs = {label: sanitize_text(str(raw[key])) for key, label in ioc_fields.items() if raw.get(key)}
if present_iocs:
parts.append("IOCs present: " + ", ".join(f"{k}={v}" for k, v in present_iocs.items()))
else:
@@ -119,14 +120,14 @@ def _build_alert_context(state: InvestigationState) -> str:
techniques = raw.get("mitre_techniques", [])
if techniques:
- parts.append(f"MITRE Techniques: {', '.join(techniques)}")
+ parts.append(f"MITRE Techniques: {', '.join(sanitize_text(str(t)) for t in techniques)}")
extra_keys = {k for k in raw if k not in {"severity", "risk_score", "mitre_techniques", *ioc_fields}}
if extra_keys:
extras = {k: raw[k] for k in sorted(extra_keys)[:10]}
parts.append("Additional fields (summary, not raw JSON):\n" + format_extra_fields_for_llm(extras))
- return "\n".join(parts)
+ return wrap_untrusted("\n".join(parts), label="alert_telemetry")
def _parse_llm_response(text: str) -> dict[str, Any]:
diff --git a/services/agents/app/agents/cloud_agent.py b/services/agents/app/agents/cloud_agent.py
index 347b802b2..01b72f14d 100644
--- a/services/agents/app/agents/cloud_agent.py
+++ b/services/agents/app/agents/cloud_agent.py
@@ -19,6 +19,7 @@ from langchain_core.messages import HumanMessage, SystemMessage
from langchain_openai import ChatOpenAI
from app.context import ContextBundle
+from app.investigator.prompt_sanitizer import sanitize_text, wrap_untrusted
from app.llm import safe_ainvoke
from app.models.state import AgentStatus, InvestigationState
from app.prompt_serialization import format_extra_fields_for_llm, summarize_structure_for_llm
@@ -63,8 +64,8 @@ def _build_cloud_context(state: InvestigationState) -> str:
"""Serialise alert data into a cloud-focused analysis prompt."""
raw = state.raw_alert
parts = [
- f"Alert Summary: {state.alert_summary}",
- f"Severity: {raw.get('severity', 'unknown')}",
+ f"Alert Summary: {sanitize_text(state.alert_summary)}",
+ f"Severity: {sanitize_text(str(raw.get('severity', 'unknown')))}",
]
cloud_fields = {
@@ -84,10 +85,10 @@ def _build_cloud_context(state: InvestigationState) -> str:
}
for key, label in cloud_fields.items():
if raw.get(key):
- parts.append(f"{label}: {raw[key]}")
+ parts.append(f"{label}: {sanitize_text(str(raw[key]))}")
if raw.get("bucket_acl") or raw.get("bucket_policy"):
- parts.append(f"Bucket ACL: {raw.get('bucket_acl', 'N/A')}")
+ parts.append(f"Bucket ACL: {sanitize_text(str(raw.get('bucket_acl', 'N/A')))}")
if raw.get("bucket_policy"):
parts.append(
"Bucket Policy:\n" + summarize_structure_for_llm(raw["bucket_policy"], label="bucket_policy", max_lines=28, max_depth=3)
@@ -110,7 +111,7 @@ def _build_cloud_context(state: InvestigationState) -> str:
parts.append(f"Public access: {raw['is_public']}")
if raw.get("finding_type"):
- parts.append(f"Finding type: {raw['finding_type']}")
+ parts.append(f"Finding type: {sanitize_text(str(raw['finding_type']))}")
extra_keys = {
k
@@ -134,7 +135,7 @@ def _build_cloud_context(state: InvestigationState) -> str:
extras = {k: raw[k] for k in sorted(extra_keys)[:8]}
parts.append("Additional fields:\n" + format_extra_fields_for_llm(extras, max_keys=8))
- return "\n".join(parts)
+ return wrap_untrusted("\n".join(parts), label="cloud_telemetry")
def _parse_response(text: str) -> dict[str, Any]:
diff --git a/services/agents/app/agents/identity_agent.py b/services/agents/app/agents/identity_agent.py
index b9242dacf..cb43122ef 100644
--- a/services/agents/app/agents/identity_agent.py
+++ b/services/agents/app/agents/identity_agent.py
@@ -18,6 +18,7 @@ from langchain_core.messages import HumanMessage, SystemMessage
from langchain_openai import ChatOpenAI
from app.context import ContextBundle
+from app.investigator.prompt_sanitizer import sanitize_text, wrap_untrusted
from app.llm import safe_ainvoke
from app.models.state import AgentStatus, InvestigationState
from app.prompt_serialization import format_extra_fields_for_llm, summarize_structure_for_llm
@@ -62,8 +63,8 @@ def _build_identity_context(state: InvestigationState) -> str:
"""Serialise alert data into an identity-focused analysis prompt."""
raw = state.raw_alert
parts = [
- f"Alert Summary: {state.alert_summary}",
- f"Severity: {raw.get('severity', 'unknown')}",
+ f"Alert Summary: {sanitize_text(state.alert_summary)}",
+ f"Severity: {sanitize_text(str(raw.get('severity', 'unknown')))}",
]
identity_fields = {
@@ -81,7 +82,7 @@ def _build_identity_context(state: InvestigationState) -> str:
}
for key, label in identity_fields.items():
if raw.get(key):
- parts.append(f"{label}: {raw[key]}")
+ parts.append(f"{label}: {sanitize_text(str(raw[key]))}")
if raw.get("login_attempts"):
parts.append(
@@ -130,7 +131,7 @@ def _build_identity_context(state: InvestigationState) -> str:
extras = {k: raw[k] for k in sorted(extra_keys)[:8]}
parts.append("Additional fields:\n" + format_extra_fields_for_llm(extras, max_keys=8))
- return "\n".join(parts)
+ return wrap_untrusted("\n".join(parts), label="identity_telemetry")
def _parse_response(text: str) -> dict[str, Any]:
diff --git a/services/agents/app/agents/insider_threat_agent.py b/services/agents/app/agents/insider_threat_agent.py
index 4de3e659a..340136a2d 100644
--- a/services/agents/app/agents/insider_threat_agent.py
+++ b/services/agents/app/agents/insider_threat_agent.py
@@ -19,6 +19,7 @@ from langchain_core.messages import HumanMessage, SystemMessage
from langchain_openai import ChatOpenAI
from app.context import ContextBundle
+from app.investigator.prompt_sanitizer import sanitize_text, wrap_untrusted
from app.llm import safe_ainvoke
from app.models.state import AgentStatus, InvestigationState
from app.prompt_serialization import format_extra_fields_for_llm, summarize_structure_for_llm
@@ -66,8 +67,8 @@ def _build_insider_context(state: InvestigationState) -> str:
"""Serialise alert data into an insider-threat focused analysis prompt."""
raw = state.raw_alert
parts = [
- f"Alert Summary: {state.alert_summary}",
- f"Severity: {raw.get('severity', 'unknown')}",
+ f"Alert Summary: {sanitize_text(state.alert_summary)}",
+ f"Severity: {sanitize_text(str(raw.get('severity', 'unknown')))}",
]
user_fields = {
@@ -82,7 +83,7 @@ def _build_insider_context(state: InvestigationState) -> str:
}
for key, label in user_fields.items():
if raw.get(key):
- parts.append(f"{label}: {raw[key]}")
+ parts.append(f"{label}: {sanitize_text(str(raw[key]))}")
activity_fields = {
"action": "Action",
@@ -97,7 +98,7 @@ def _build_insider_context(state: InvestigationState) -> str:
}
for key, label in activity_fields.items():
if raw.get(key):
- parts.append(f"{label}: {raw[key]}")
+ parts.append(f"{label}: {sanitize_text(str(raw[key]))}")
if raw.get("access_time") or raw.get("timestamp"):
parts.append(f"Access time: {raw.get('access_time') or raw.get('timestamp')}")
@@ -150,7 +151,7 @@ def _build_insider_context(state: InvestigationState) -> str:
extras = {k: raw[k] for k in sorted(extra_keys)[:8]}
parts.append("Additional fields:\n" + format_extra_fields_for_llm(extras, max_keys=8))
- return "\n".join(parts)
+ return wrap_untrusted("\n".join(parts), label="insider_threat_telemetry")
def _parse_response(text: str) -> dict[str, Any]:
diff --git a/services/agents/app/agents/phishing_agent.py b/services/agents/app/agents/phishing_agent.py
index c8fa991c2..e03f9f70f 100644
--- a/services/agents/app/agents/phishing_agent.py
+++ b/services/agents/app/agents/phishing_agent.py
@@ -19,6 +19,7 @@ from langchain_core.messages import HumanMessage, SystemMessage
from langchain_openai import ChatOpenAI
from app.context import ContextBundle
+from app.investigator.prompt_sanitizer import sanitize_text, wrap_untrusted
from app.llm import safe_ainvoke
from app.models.state import AgentStatus, InvestigationState
from app.prompt_serialization import format_extra_fields_for_llm, summarize_structure_for_llm
@@ -55,8 +56,8 @@ def _build_phishing_context(state: InvestigationState) -> str:
"""Serialise alert data into a phishing-focused analysis prompt."""
raw = state.raw_alert
parts = [
- f"Alert Summary: {state.alert_summary}",
- f"Severity: {raw.get('severity', 'unknown')}",
+ f"Alert Summary: {sanitize_text(state.alert_summary)}",
+ f"Severity: {sanitize_text(str(raw.get('severity', 'unknown')))}",
]
email_fields = {
@@ -69,14 +70,14 @@ def _build_phishing_context(state: InvestigationState) -> str:
}
for key, label in email_fields.items():
if raw.get(key):
- parts.append(f"{label}: {raw[key]}")
+ parts.append(f"{label}: {sanitize_text(str(raw[key]))}")
if raw.get("urls"):
parts.append("URLs found:\n" + summarize_structure_for_llm(raw["urls"], label="urls", max_lines=24, max_depth=2))
if raw.get("url"):
- parts.append(f"Primary URL: {raw['url']}")
+ parts.append(f"Primary URL: {sanitize_text(str(raw['url']))}")
if raw.get("domain"):
- parts.append(f"Domain: {raw['domain']}")
+ parts.append(f"Domain: {sanitize_text(str(raw['domain']))}")
if raw.get("attachment_hashes"):
parts.append(
@@ -84,7 +85,7 @@ def _build_phishing_context(state: InvestigationState) -> str:
+ summarize_structure_for_llm(raw["attachment_hashes"], label="attachment_hashes", max_lines=16, max_depth=1)
)
if raw.get("file_hash"):
- parts.append(f"File hash: {raw['file_hash']}")
+ parts.append(f"File hash: {sanitize_text(str(raw['file_hash']))}")
auth_results = {
"spf_result": "SPF",
@@ -94,12 +95,12 @@ def _build_phishing_context(state: InvestigationState) -> str:
auth_parts = []
for key, label in auth_results.items():
if raw.get(key):
- auth_parts.append(f"{label}={raw[key]}")
+ auth_parts.append(f"{label}={sanitize_text(str(raw[key]))}")
if auth_parts:
parts.append(f"Email auth: {', '.join(auth_parts)}")
if raw.get("body_snippet"):
- parts.append(f"Body snippet: {raw['body_snippet'][:500]}")
+ parts.append(f"Body snippet: {sanitize_text(str(raw['body_snippet']), max_len=500)}")
extra_keys = {
k
@@ -124,7 +125,7 @@ def _build_phishing_context(state: InvestigationState) -> str:
extras = {k: raw[k] for k in sorted(extra_keys)[:8]}
parts.append("Additional fields:\n" + format_extra_fields_for_llm(extras, max_keys=8))
- return "\n".join(parts)
+ return wrap_untrusted("\n".join(parts), label="phishing_telemetry")
def _parse_response(text: str) -> dict[str, Any]: