skill

dcl-skills

Security & compliance skills for AI agents (jailbreak defense, PII/secret redaction, hallucination checks) with a DeFi-specialized suite — free instruction-only mode or paid on-chain-verified mode via

Fronesis-Labs/dcl-skillscommunitylistedunreviewed

Use skill

Set up

Skills are installed by placing their folder where your agent loads skills.

Download ZIP from GitHub Open the repository

The download comes from github.com/Fronesis-Labs/dcl-skills itself, not from our server, so what you get is exactly what its author published.

Install instructions come from the README

Skills do not use an MCP config. Follow the steps in the README below.

README

synced Oct 28, 2026

DCL Skills

Don't trust the agent. Trust the proof.

Compliance and security skills for AI agents — hallucination detection, PII redaction, prompt-injection defense, secret-leak detection, and a DeFi-specialized variant set for agents that touch wallets and on-chain actions.

Published on ClawHub under @daririnch.

Why This Exists

An agent that can act autonomously is also an agent that can be tricked, drift off its instructions, or leak something it shouldn't — silently, with no record. These skills give an agent (or the pipeline running it) a way to check itself before acting and to leave a verifiable trail of what it checked. Most skills work two ways: a free, instruction-only checklist you can run yourself right now, and an optional live audit backed by the DCL Trust Oracle for a cryptographically verifiable verdict.

Structure

dcl-skills/
├── skills/
│   ├── core/      # general-purpose agent-safety skills
│   └── crypto/     # DeFi-specialized variants (wallets, trades, MEV, on-chain provenance)
├── docs/          # pipeline design, security model, whitepaper
├── tests/golden/  # golden-set eval cases per skill
└── .github/workflows/lint-skills.yml

Skills

Core

Skill	Purpose
`dcl-policy-enforcer`	Enforce compliance policy on agent actions before execution
`dcl-sentinel-trace`	Detect & redact PII in agent outputs
`dcl-prompt-firewall`	Detect & block prompt injection
`dcl-secret-leak-detector`	Detect leaked secrets/credentials
`dcl-provenance-tracker`	Verify skill version integrity / detect supply-chain drift after updates
`dcl-skill-auditor`	Pre-install static security scanner for ClawHub skills
`dcl-semantic-drift-guard`	Detect hallucination / context drift vs grounding sources
`dcl-pipeline`	Meta-skill orchestrating the above as one pipeline

Crypto

Skill	Purpose
`dcl-prompt-firewall-crypto`	Injection defense specialized for DeFi contexts
`dcl-wallet-guardian`	Validate wallet-affecting actions before execution
`dcl-trade-verifier`	Verify trade/swap parameters before execution
`dcl-mev-compliance`	Check MEV exposure / compliance before broadcast
`dcl-provenance-tracker-crypto`	Verify provenance of on-chain data/contracts referenced
`dcl-semantic-drift-crypto`	Crypto-specific hallucination detection (token/contract claims)
`dcl-output-sanitizer`	Sanitize agent output before display/execution
`dcl-crypto-commit`	Final compliance checkpoint before irreversible on-chain action

Free by Default, Verifiable if You Need It

Every skill runs standalone as a free, instruction-only checklist — no account, no network call, no dependency on Fronesis Labs infrastructure. Most also offer an optional live check against the DCL Trust Oracle for a cryptographically sealed, tamper-evident verdict instead of a self-reported one, metered per call (settled automatically via x402, USDC on Base) so there's no subscription to manage for occasional or high-volume use. dcl-crypto-commit is the one skill with no free equivalent, since its whole purpose is writing to the on-chain audit chain. dcl-provenance-tracker and dcl-provenance-tracker-crypto are the opposite case — 100% instruction-only by design, with only an optional on-chain cross-check of a prior record. See each skill's SKILL.md for details.

Status

All 16 SKILL.md files under skills/ (8 core + 8 crypto) have been migrated from ClawHub and are complete — see docs/pipeline.md for the per-skill migration log, including notes on which MCP tool names/prices were inferred by analogy rather than confirmed from a source doc (mainly the crypto suite, which started from thin ClawHub listings).

License

This repo: Apache License 2.0 — see LICENSE.
ClawHub listings: platform-enforced MIT-0, cannot be overridden per skill.

See docs/licensing.md for the full breakdown, including an open item on x402/paid-audit wording vs ClawHub's no-pricing-metadata rule.