feat(core): give Waggle its own database, apart from memory (P1.5)
Waggle kept its own tables in the memory database. Before the memory
engine can be replaced by mem0 they need a home of their own.
- add WaggleDB (<dataDir>/waggle.db). Automations, notifications, held
actions, skill hashes, thumbs feedback, cost tracking, the profile card
and awareness items live there. Their stores take a WaggleDB, so passing
a memory database is a type error.
- the profile card and awareness items are now per scope ('personal' or
'workspace:<id>') in one table each, instead of one table per .mind file
- on the first start, copy these tables from personal.mind and every
workspace.mind (importLegacyMindData). The legacy files are only read,
each scope is copied once and a table only while its target is empty.
- copy the logger, injection scanner, external-memory ingress guard,
WorkspaceManager and the embedders from @waggle/hive-mind-core into
@waggle/core, which now re-exports its own copies
- point the server, agent orchestrator and CLI at waggle.db; update the
tests that seeded these tables in a memory database
ADR-0007 records the layout and the licensing question about the copied
files.
This commit is contained in:
85
packages/core/src/injection-scanner.ts
Normal file
85
packages/core/src/injection-scanner.ts
Normal file
@@ -0,0 +1,85 @@
|
||||
/**
|
||||
* Prompt-injection scanner — pattern-based detection for role-override,
|
||||
* prompt-extraction, and instruction-injection attempts.
|
||||
*
|
||||
* Historically lived in `@waggle/agent`. Moved to `@waggle/core` so the
|
||||
* harvest pipeline (which ingests untrusted external conversation exports)
|
||||
* can call it without a cross-package dependency. The `@waggle/agent`
|
||||
* version is now a thin re-export for backward compatibility.
|
||||
*
|
||||
* Three pattern sets:
|
||||
* - ROLE_OVERRIDE: "ignore previous instructions", "you are now…", memory-wipe attempts
|
||||
* - PROMPT_EXTRACTION: "print your system prompt", "reveal your instructions"
|
||||
* - INSTRUCTION_INJECTION: fake "IMPORTANT:", "SYSTEM:", "[INST]" authority markers
|
||||
*/
|
||||
|
||||
export interface ScanResult {
|
||||
safe: boolean;
|
||||
score: number;
|
||||
flags: string[];
|
||||
}
|
||||
|
||||
const ROLE_OVERRIDE_PATTERNS = [
|
||||
/ignore\s+(all\s+)?previous\s+(instructions|prompts|rules)/i,
|
||||
/you\s+are\s+now\s+/i,
|
||||
/new\s+instructions?\s*:/i,
|
||||
/forget\s+(everything|all|your)\s+(you|instructions|rules)/i,
|
||||
/override\s+(your|the|all)\s+(instructions|rules|prompt)/i,
|
||||
/disregard\s+(your|the|all|previous)\s+(instructions|rules|prompt)/i,
|
||||
/ignoriere\s+alle/i,
|
||||
/ignora\s+todas/i,
|
||||
/ignorez\s+toutes/i,
|
||||
// F8: Memory wipe — "forget everything", "erase all context", etc.
|
||||
/(?:forget|erase|clear|wipe|reset)\s+(?:everything|conversation|memory|context|history|all)/i,
|
||||
// F8: Role override — "from now on you are", "pretend you are", "your new role is"
|
||||
/(?:from\s+now\s+on\s+you\s+are|act\s+as\s+if\s+you\s+are|pretend\s+you\s+are|your\s+new\s+role\s+is)/i,
|
||||
];
|
||||
|
||||
const PROMPT_EXTRACTION_PATTERNS = [
|
||||
/print\s+(your|the)\s+system\s+prompt/i,
|
||||
/show\s+(me\s+)?(your|the)\s+(system\s+)?prompt/i,
|
||||
/what\s+(is|are)\s+your\s+(system\s+)?(prompt|instructions|rules)/i,
|
||||
/repeat\s+(your|the)\s+(system\s+)?(prompt|instructions)/i,
|
||||
/output\s+(your|the)\s+(system\s+)?prompt\s+verbatim/i,
|
||||
// F8: Instruction disclosure — "reveal your instructions", "disclose your prompt", etc.
|
||||
/(?:reveal|disclose|show|display|print|output|dump)\s+(?:your\s+)?(?:instructions|system\s+prompt|rules|guidelines|configuration|training)/i,
|
||||
];
|
||||
|
||||
const INSTRUCTION_INJECTION_PATTERNS = [
|
||||
/IMPORTANT\s*:\s*(ignore|disregard|forget|override)/i,
|
||||
/SYSTEM\s*:\s*/i,
|
||||
/\[INST\]/i,
|
||||
/<<SYS>>/i,
|
||||
/\bASSISTANT\s*:\s*/i,
|
||||
/BEGIN\s+NEW\s+INSTRUCTIONS/i,
|
||||
// F8: Authority claims — fake system/admin messages
|
||||
/(?:system\s+message|admin\s+message|admin\s+override|elevated\s+privileges|root\s+access|operator\s+mode|maintenance\s+mode|debug\s+mode)/i,
|
||||
];
|
||||
|
||||
export function scanForInjection(
|
||||
text: string,
|
||||
context: 'user_input' | 'tool_output' = 'user_input'
|
||||
): ScanResult {
|
||||
const flags: string[] = [];
|
||||
let score = 0;
|
||||
|
||||
for (const pattern of ROLE_OVERRIDE_PATTERNS) {
|
||||
if (pattern.test(text)) { flags.push('role_override'); score += 0.5; break; }
|
||||
}
|
||||
|
||||
for (const pattern of PROMPT_EXTRACTION_PATTERNS) {
|
||||
if (pattern.test(text)) { flags.push('prompt_extraction'); score += 0.4; break; }
|
||||
}
|
||||
|
||||
for (const pattern of INSTRUCTION_INJECTION_PATTERNS) {
|
||||
if (pattern.test(text)) {
|
||||
flags.push('instruction_injection');
|
||||
score += context === 'tool_output' ? 0.6 : 0.3;
|
||||
break;
|
||||
}
|
||||
}
|
||||
|
||||
score = Math.min(score, 1.0);
|
||||
|
||||
return { safe: score < 0.3, score, flags };
|
||||
}
|
||||
Reference in New Issue
Block a user