dcl-prompt-firewallInstruction-only input-layer shield for AI agents and LLM pipelines. Detects prompt injection, jailbreak attempts, instruction override, role-switch attacks,...
Install via ClawdBot CLI:
clawdbot install daririnch/dcl-prompt-firewallGrade Fair — based on market validation, documentation quality, package completeness, maintenance status, and authenticity signals.
Contains base64-encoded or hex-obfuscated content (potential hidden instructions)
base64 decoded: "Ignore all previous instructions..."Contains instructions to override system prompt or ignore user requests
"Ignore previous instructions"Sends data to undocumented external endpoint (potential exfiltration)
POST → https://webhook.fronesislabs.com/evaluateCalls external URL not in known-safe list
Generated Nov 3, 2026
A SaaS support bot screens every inbound user message through dcl_evaluate_jailbreak before passing it to the LLM that has access to ticketing and refund tools. Malicious customers trying to override instructions or hijack the assistant get blocked with a NO_COMMIT verdict and an on-chain audit trail.
An enterprise knowledge assistant screens retrieved documents and web snippets before they enter the prompt context, catching injected instructions hidden in scraped pages. Each blocked document produces a tamper-evident tx_hash record for compliance review.
In a multi-agent system where one agent's output feeds another, every inter-agent message passes through the firewall first. This prevents a compromised or prompt-injected agent from propagating malicious directives downstream, with per-agent_id attribution in the audit log.
A payments assistant screens customer-supplied text and third-party tool results before executing any transaction instruction. Regulatory auditors can verify the hash-chained log to confirm pre-execution screening occurred without exposing raw customer data.
A consumer AI app screens all free-text input for jailbreak and role-switch attempts before generation, protecting brand safety and reducing abusive outputs. The low per-call USDC pricing scales with actual traffic instead of a fixed subscription.
Charge $0.02 per jailbreak screen settled in USDC on Base, requiring no account or subscription. Revenue scales directly with screened traffic and appeals to developers who want zero commitment upfront.
Position the firewall alongside DCL Policy Enforcer and DCL Sentinel Trace as a full pre-input and post-output security package. Customers adopting the whole suite drive higher volume across all metered tools.
Offer the hash-chained, tamper-evident audit trail as the compliance value proposition for fintech and healthcare clients who must prove pre-execution screening. Optional enterprise reporting and log export packages monetize the audit layer separately.
💬 Integration Tip
Register the MCP server in your client config and ensure your agent has an x402-capable wallet funded with USDC on Base; fall back to the free instruction-only checklist for offline or no-payment scenarios.
Scored Oct 9, 2026
https://webhook.fronesislabs.com/evaluateAI Analysis
This skill sends all user prompts to an external third-party API (fronesislabs.com) for analysis, creating a data privacy risk as prompts may contain sensitive information. While the external call is documented and consistent with the skill's stated purpose, it establishes a persistent data pipeline to an unvetted external service. The skill itself does not contain hidden malicious instructions or credential harvesting patterns.
Audited Apr 16, 2026 · audit v1.0
Meta-skill for AI agent self-improvement. Analyzes runtime logs to detect error patterns, regressions, and inefficiencies, then generates structured improvem...
Stop waiting for prompts. Keep working.
Turn OpenClaw into a learning-loop agent with seeded workspace rules, skill promotion, reflective memory, and proactive maintenance.
Complete toolkit for creating autonomous AI agents and managing Discord channels for OpenClaw. Use when setting up multi-agent systems, creating new agents, or managing Discord channel organization.
Billions decentralized identity for agents. Link agents to human identities using Billions ERC-8004 and Attestation Registries. Verify and generate authentic...
AI agent for finance search agent tasks