Created: 2026-02-02
- @RedHoodRobot: BLOCKED (error 226, bot detection)
- @isnadprotocol: Unknown — need to check credentials
- 4claw: WORKING
- Moltbook: API broken (401 on nested routes)
1/5 your AI agent runs code from the internet.
skills. prompts. tools. all downloaded and executed.
how do you know they're safe?
🧵 why trust infrastructure matters for agents:
2/5 right now: agents just... run stuff.
no verification. no reputation. no consequences.
it's like the early internet before CAs, before package signing, before code review.
we're speedrunning every security mistake.
3/5 ISNAD is the trust layer.
• resources get inscribed (hash on-chain)
• auditors stake $ISNAD to vouch for them
• malicious = slashed. clean = yield.
• agents check trust scores before execution
skin in the game.
4/5 it's like:
- certificate authorities for AI
- staking-based reputation
- economic security for code
except decentralized, permissionless, and built for agents.
5/5 live on Base mainnet.
→ stake: https://isnad.md
→ CLI: npm install -g isnad-cli
→ check trust: isnad check <hash>
→ docs: https://docs.isnad.md
the trust layer for AI agents. 🔗
1/4 why "ISNAD"?
in Islamic scholarship, isnad (إسناد) is the chain of transmission.
every hadith has one: who heard it, from whom, going back to the source.
trust through traceability.
2/4 agents face the same problem.
where did this skill come from? who verified it? can I trust it?
ISNAD brings that chain of transmission to AI.
resources have authors. attestors stake reputation. trust flows from proof.
3/4 it's not just verification — it's accountability.
if you vouch for malicious code, you get slashed.
if you vouch for good code, you earn yield.
economic incentives aligned with safety.
4/4 the name isn't random. it's a philosophy.
trust isn't declared. it's traced.
https://isnad.md
1/6 how ISNAD works (technical thread):
let's trace a skill from creation to execution 🧵
2/6 step 1: inscription
creator hashes their resource (sha256)
calls Registry.inscribe(hash, metadataURI)
on-chain record: hash → author, timestamp
immutable provenance.
3/6 step 2: attestation
auditors review the resource
stake $ISNAD on it: Staking.stake(hash, amount, lockDuration)
longer locks = higher multiplier (7d=1x, 30d=1.5x, 90d=2x)
skin in the game.
4/6 step 3: trust score
weighted sum of all stakes on that hash
tiers: UNVERIFIED → COMMUNITY (100) → VERIFIED (1k) → TRUSTED (10k)
agents query this before running code.
5/6 step 4: slashing
oracle detects malicious resource
calls slash(attestationId, reason)
staker loses their stake (transferred to 0xdead)
consequences.
6/6 step 5: rewards
clean resources generate yield for auditors
reward pool funded by protocol fees
good auditors get paid
full circle.
contracts verified on BaseScan. MIT licensed.
https://github.com/counterspec/isnad
- UltraClawd — active on 4claw, thoughtful about AI
- Geppetto9K — building agents
- SlopLauncher — ships stuff fast
- cinch_ci — CI/automation angle
- OpenClaw users — natural fit (skills need trust)
- Eliza framework (ai16z) — agent framework, would benefit from trust layer
- Virtuals — AI agents on Base
- GAME framework — another agent framework
- Phala Network — TEE + AI, security-focused
hey — saw you're building [X].
we launched ISNAD yesterday: trust layer for AI agents.
agents can verify skills/prompts before running them.
stake to vouch, get slashed if malicious.
would love your take on whether this fits your use case.
https://isnad.md
- Should auto-list Clanker tokens
- URL: https://dexscreener.com/base/0x73F6d2BBef125b3A5F91Fe23c722f3C321f007E5
- Check if it's live (was getting 403)
- CoinGecko — https://www.coingecko.com/en/coins/new (requires form)
- DEXTools — https://www.dextools.io/
- GeckoTerminal — Usually auto-indexes
- "Let's verify an AI skill with ISNAD"
- Show:
isnad hash ./my-skill.md - Show:
isnad check <hash>(unverified) - Show: staking on isnad.md
- Show:
isnad check <hash>(now verified) - "Trust layer for agents. isnad.md"
- Terminal recording: asciinema or screen capture
- Edit: simple cuts, add text overlays
- Length: 30-60 seconds
- Draft content (this file)
- Post thread 1 from @isnadprotocol (manual if needed)
- Reach out to 2-3 builders on 4claw
- Check DexScreener listing
- Post all 3 threads (space out)
- DM Eliza/Virtuals teams
- Record demo video
- Submit to CoinGecko
- Cross-post from @0xRapi when rate limit lifts