IdentArk @identArk

AI-native secrets management for autonomous agents. Capabilities, not credentials. Control access. Audit every action. Approve risk. Joined September 2026

Tweets

17
Followers

5
Following

50
Likes

13

IdentArk @identArk

2 days ago

At the sink, not the client. In our CLI, human output gets controls stripped, then redacted, then escaped. --json mode can't leak them: the encoder turns ESC into \u001b, so nothing raw reaches the tty. The proxy can't know if bytes are headed for a terminal, a browser or a JSON parser, and each needs different escaping. Careful with logging raw bytes, though. Raw is where the key lives. Log the redacted form plus a hash of the raw, or keep raw behind the same access controls as your secret store.

0 0 1 8 0

View Details

IdentArk @identArk

2 days ago

Your MCP tool returns a string. Your CLI prints it. That string came from a server you don't control, and it lands in a terminal that interprets control characters. Two bugs hide in that one sentence. 🧵

2 0 0 13 0

View Details

IdentArk @identArk

2 days ago

IdentArk — your agent never holds the key. identark.io

0 0 0 0 0

View Details

IdentArk @identArk

2 days ago

None of this is clever. It's a handful of operations in the right order, in one helper, called at every point where a remote value reaches the screen. The bug was never the regex. It's running a regex over input you haven't normalised yet, and assuming a string is just a string.

1 0 0 0 0

View Details

IdentArk @identArk

5 days ago

@Pontifex The Pope knows about Ontology

0 0 0 17 0

View Details

IdentArk @identArk

2 weeks ago

Before: your agent knows all your passwords. After: it doesn't need to - and everything it does is on the record. Same workflow. Same output. Different blast radius.

0 1 1 22 0

View Details

IdentArk @identArk

2 weeks ago

Two items on that list want a different home than the rest. Permissions and pending actions can't live in the agent's own state - anything the agent can read, injected text can influence. A pending approval the agent can see is one it can talk itself into. Progress and retries are fine in-process. Authority isn't.

0 0 0 11 0

View Details

IdentArk @identArk

2 weeks ago

@liorb_d The split that usually gets missed: separating "use this tool" from "reconfigure this tool." We keep those as distinct scopes - an agent key can invoke an MCP server all day but can't re-point it at a different endpoint. Bundle them and the blast radius doubles quietly.

0 0 0 39 0

View Details

IdentArk @identArk

3 weeks ago

Open-source opportunity most builders are sleeping on: You don’t need a PhD to move the needle on secure AI agents. What actually creates leverage in 2026 (and helps IdentArk): • Take any agent paper / framework and re-implement it against the AgentGateway Protocol • Write a clean evaluation or stress test for zero-secret execution, cost tracking, or session isolation • Ship a small, reusable integration, example, or gateway extension people actually adopt • Document what broke when you tried to run agents without secrets in the process • Talk to teams already deploying agents before you “do research” Proof of secure execution > demos. Reusable protocol tooling > one-off projects. Clarity on the security/reliability gap is what gets amplified. If you’re a builder this week: Pick one agent system. Port it to the open-source IdentArk SDK. Publish the repo + the failure modes you hit. That’s the contribution that compounds for you and for the protocol.