jlugo63

The consequences to our basic character is a fixed result.

Agentic AI Engineer | AWS, Databricks

[jlugo63 @Jermamerma]

I built an open-source governance runtime for AI agents.
Every tool call goes through a policy engine before it executes. If the agent tries something dangerous, it gets blocked mechanically, not by asking nicely.

Tested against OpenClaw 2026.2.15. It works.
🔗 https://t.co/tRoB98g9lT

[jlugo63 @Jermamerma]

@asaio87 @DylanLister88 Can't hate man. The SWE role has changed. New world for this profession now. Sink or swim.

[jlugo63 @Jermamerma]

Governance for agentic AI will absolutely exist.

The only open questions are: Who builds it? What form it takes? When it becomes demanded instead of optional.

Why isn’t everyone building this?

Because:

Most agents today are still low-stakes

The damage hasn’t been large-scale yet

People optimize for speed, not safety

Governance feels like friction… until it isn’t

Security always looks unnecessary until the first serious incident.

https://t.co/zyWUaML8gq

[jlugo63 @Jermamerma]

Apache 2.0. Run it yourself.
https://t.co/G014HcRveq

[jlugo63 @Jermamerma]

For OpenClaw: a TypeScript plugin registers a before_tool_call hook. The agent literally cannot execute without gateway approval. For any other framework: Python SDK, five lines. New in today's build: approved commands run inside a sandboxed Docker container. Every execution produces a tamper-evident evidence packet logged to the audit chain.

[jlugo63 @Jermamerma]

Every decision is written to a hash-chained audit ledger in PostgreSQL.
Each event's SHA-256 hash includes the previous event's hash. 553 events verified back to genesis. Tamper with one record and the entire chain breaks.
This isn't a log file. It's a cryptographic proof of what your agent did and why.

[Joaki @itsjoaki]

everyone on my timeline is running billion dollar businesses with 100 openclaw agents

wild how not a single one can show what the agents actually produced

[jlugo63 @Jermamerma]

Here's what it looks like when an OpenClaw agent tries to write to a protected file.
The hook intercepts the tool call. The policy engine evaluates it. §I.2 Authority Decoupling fires. Risk score: 0.9. DENIED.
The agent never touches the file.

[jlugo63 @Jermamerma]

Gavel sits between the agent and execution.
Every action shell commands, file writes, API calls gets proposed to a policy engine before it runs.

The engine evaluates it against a written constitution and returns: → APPROVED (proceed) → DENIED (blocked, violations cited) → ESCALATED (human must approve)

#CyberSecurity #OpenClaw

[jlugo63 @Jermamerma]

OpenClaw has 195K GitHub stars. Fortune, Cisco, Kaspersky, and VentureBeat have all published pieces in the last two weeks saying the same thing: AI agents with shell access and no guardrails are a security nightmare.
The agent can rm -rf your machine, exfiltrate data, or get hijacked via prompt injection. And there's almost nothing in the ecosystem to stop it.

#OpenClaw

[jlugo63 @Jermamerma]

@asynchronous_x Skill issue

[jlugo63 @Jermamerma]

@mazeincoding You don't know how to program your own browser from scratch? Lmaooooo NOOB

[jlugo63 @Jermamerma]

@ajackhorse @SeanJohns1972 @RedWave_Press I feel like any path you took on that spaghetti thought it was always going to end in. "Possible foreign nations."

[Nic Cruz Patane @niccruzpatane]

Just leaving this here as a reminder

[kira 👾 @kirawontmiss]

The Mathematics and Physics y’all said don’t matter in life

[Dudes Posting Their W’s @DudespostingWs]

Dudes will see this and say hell yea

[Gabriel @GabeZZOZZ]

👀 I never saw Putin in military uniform

Ukraine is FUCKED

[jlugo63 @Jermamerma]

@elonmusk King autist trying to shape my opinion using a fucking meme. Lmao