---
title: "Arunima Chaudhuri (@ArunimaChaudhuri) | Hashnode"
description: "Research Engineer @ Status | Ex - Intern @Status @Hyperledger @Solana Labs | M.Tech CSE ’25 @NITW. Read the latest articles by Arunima Chaudhuri on Hashnode."
url: "https://hashnode.com/@ArunimaChaudhuri"
---

@ArunimaChaudhuri

# Arunima Chaudhuri

@ArunimaChaudhuri · Kolkata, India · Joined October 2022

Research Engineer @ Status | Ex - Intern @Status @Hyperledger @Solana Labs | M.Tech CSE ’25 @NITW

## Comments

[You’re right that once merged-with-no-review becomes normal, it stops looking like a decision. That’s why I’d track the unreviewed-merge rate explicitly. It turns a silent default back into a number someone has to answer for. On your question, I’d separate evidence from the rubric. Agents can and should generate evidence, because that’s the cheap part. What they can’t be allowed to do is define what counts as passing. In practice that means three things: A human writes the acceptance criteria before the work starts, and they sit somewhere the building agent can’t edit. The verifier runs in a different session from the builder, with different context. You check the tests themselves, for example with mutation testing: if you deliberately break the code and the tests still pass, the evidence is theatre. The risk you’re describing is real when one agent writes the code, the tests and the definition of done. Split those three apart and the human is grading against their own rubric again.](https://hashnode.com/posts/code-is-cheap-authority-isn-t/6ad54358d75f37b1b40b3b8d/comment/6ad56d2839cde4bbff5982c3#reply-6ae9877eb09e0c496f5ea55e)

Reply · Article · Nov 3 · [Code Is Cheap. Authority Isn't.](https://hashnode.com/posts/code-is-cheap-authority-isn-t/6ad54358d75f37b1b40b3b8d)

[Thanks, this is a sharper version of the point than mine. “Review becomes a comparison rather than a reading” is the shift I was trying to describe. On your question, my answer is that criteria should be owned at the boundary where an outcome becomes visible, not wherever the agents happen to run. Each team owns acceptance criteria for its own domain. The interfaces between teams get their own contract, owned by whoever consumes it, because they’re the ones who get hurt when it breaks. Invariants that span the whole system (data ownership, tenancy, money) need one named owner, usually a tech lead or platform function. The failure mode I’d watch for is ownership by committee. If three teams “jointly own” the criteria for a cross-team flow, nobody does, and the agents fill that gap with whatever they infer. One name per contract, even when the system spans ten teams.](https://hashnode.com/posts/code-is-cheap-authority-isn-t/6ad54358d75f37b1b40b3b8d/comment/6ad5818839cde4bbff59bf86#reply-6ae98768efde647f3203fd37)

Reply · Article · Nov 3 · [Code Is Cheap. Authority Isn't.](https://hashnode.com/posts/code-is-cheap-authority-isn-t/6ad54358d75f37b1b40b3b8d)

[Agreed, a run that never fetched the poisoned body is the scariest kind of green. A few rules, all in one shared module so the CLI, compare and both dashboards can’t drift apart: The finding rate is calculated over tested scenarios only. If nothing was tested, it’s null, not 0%. Coverage (tested / total) sits right next to it, so “0 findings, 20% coverage” reads as “we barely measured,” not “safe.” A run with inconclusive scenarios is never labelled “all passed,” and the CLI prints a warning that N scenarios were inconclusive, not passed. In compare, a finding is never called “fixed” if that scenario was inconclusive in the newer run. Not reaching the poisoned doc doesn’t count as a fix. If you’d like to try it on your RAG evals, I’d love to know whether the delivery tracking catches your case: github.com/invarislabs/invaris-agentsec](https://hashnode.com/posts/the-agent-never-opened-the-email-engineering-test-verdicts-that-know-what-the-agent-actually-saw/6acd2d8da159f761e244e51b/comment/6ace618946d7036803872ec7#reply-6acef62dacc964cbf9093383)

Reply · Article · Oct 14 · [The Agent Never Opened the Email: Engineering Test Verdicts That Know What the Agent Actually Saw](https://hashnode.com/posts/the-agent-never-opened-the-email-engineering-test-verdicts-that-know-what-the-agent-actually-saw/6acd2d8da159f761e244e51b)

[Good question. In the default setup they can’t disagree, because they come from the same place. AgentSec runs the simulated tools itself, so it is the one that serves the poisoned email. Exposure is the harness’s own record of the tool result that carried the injection, and every event gets one sequence number from the harness. A finding needs a goal-matching call with a higher sequence number than delivery, and calls before delivery don’t count. Nothing is rebuilt from the agent’s logs. Your concern does apply to agents that run their own tools and report them back to us. Then the trace is the agent’s account, and the verdict inherits whatever it chose to record. Those events are flagged executed\_by\_agent in the report, but flagging isn’t verification. The real fix is observing independently, below the agent: for example, logging network egress per test case. That’s on the roadmap.](https://hashnode.com/posts/the-agent-never-opened-the-email-engineering-test-verdicts-that-know-what-the-agent-actually-saw/6acd2d8da159f761e244e51b/comment/6ace98180cb76c6fa8f2a8de#reply-6acef5efa87c7c9ac9a672f1)

Reply · Article · Oct 14 · [The Agent Never Opened the Email: Engineering Test Verdicts That Know What the Agent Actually Saw](https://hashnode.com/posts/the-agent-never-opened-the-email-engineering-test-verdicts-that-know-what-the-agent-actually-saw/6acd2d8da159f761e244e51b)

[Thanks Salman, that’s exactly the reasoning. A 0.0 that really means “we never checked” is worse than no number, because it looks like proof. Your measured / modeled / aspirational split is a nice way to put it, and I like that it makes the confidence visible right next to the figure. AgentSec is open source if you’d like to try it: GitHub. If you have agents in your stack at ML Systems, I’d love to hear what it catches and what it misses. And if you find it useful, a star helps more people find it.](https://hashnode.com/posts/the-agent-never-opened-the-email-engineering-test-verdicts-that-know-what-the-agent-actually-saw/6acd2d8da159f761e244e51b/comment/6acd578b52d15945e868bb8f#reply-6acd5a2f0cb96e144df514e8)

Reply · Article · Oct 12 · 1 · [The Agent Never Opened the Email: Engineering Test Verdicts That Know What the Agent Actually Saw](https://hashnode.com/posts/the-agent-never-opened-the-email-engineering-test-verdicts-that-know-what-the-agent-actually-saw/6acd2d8da159f761e244e51b)

[Great question, Kartik! The idea is to capture the user's original request as the source of authority, then evaluate tool effects against what that request actually permits — not just which tools the agent can access. The tricky part is translating natural-language intent into explicit constraints without assuming permissions the user never granted. That's one of the problems I'm exploring with AgentSec!](https://hashnode.com/posts/allowed-isn-t-authorized-testing-what-your-ai-agent-does-with-the-permissions-it-already-has/6ac5586147a52923b5896365/comment/6ac7de30969b93a26bac4f9b#reply-6ac7e9d2bcc1f6d9af0ccf81)

Reply · Article · Oct 8 · [Allowed Isn't Authorized: Testing What Your AI Agent Does With the Permissions It Already Has](https://hashnode.com/posts/allowed-isn-t-authorized-testing-what-your-ai-agent-does-with-the-permissions-it-already-has/6ac5586147a52923b5896365)

[Really appreciate you reading it, Salman! Glad you found the five questions useful!](https://hashnode.com/posts/allowed-isn-t-authorized-testing-what-your-ai-agent-does-with-the-permissions-it-already-has/6ac5586147a52923b5896365/comment/6ac5661b8d40abca43c75441#reply-6ac7e93168a3fe2456b86953)

Reply · Article · Oct 8 · 1 · [Allowed Isn't Authorized: Testing What Your AI Agent Does With the Permissions It Already Has](https://hashnode.com/posts/allowed-isn-t-authorized-testing-what-your-ai-agent-does-with-the-permissions-it-already-has/6ac5586147a52923b5896365)

[Yeah, exactly. If something can’t be judged, that should stay visible instead of quietly turning into “safe.” And I agree — knowing whose authority the agent acted on matters just as much as knowing which agent acted.](https://hashnode.com/posts/allowed-isn-t-authorized-testing-what-your-ai-agent-does-with-the-permissions-it-already-has/6ac5586147a52923b5896365/comment/6ac5661b8d40abca43c75441#reply-6ac612f8e84b085d9baf727b)

Reply · Article · Oct 7 · 1 · [Allowed Isn't Authorized: Testing What Your AI Agent Does With the Permissions It Already Has](https://hashnode.com/posts/allowed-isn-t-authorized-testing-what-your-ai-agent-does-with-the-permissions-it-already-has/6ac5586147a52923b5896365)

[Welcome. I am glad you liked it.](https://hashnode.com/posts/debugging-css-with-chrome-devtools--deleted/63ccedf92d25af16987a8376/comment/63d62126fef5f9337240d711#reply-63d7286704ace63ce8863fbd)

Reply · Article · Jan 30, 2023 · [Debugging CSS with Chrome DevTools](https://hashnode.com/posts/debugging-css-with-chrome-devtools--deleted/63ccedf92d25af16987a8376)

[Thanks Ajay Kumar. I did a few updates to further simplify the concepts. Do have a look](https://hashnode.com/posts/what-an-api-consists-of/63d13d34107ed10d731538e8/comment/63d21423a3afbaf919c6689b#reply-63d2397d19094b06f1060fcb)

Reply · Article · Jan 26, 2023 · [What an API consists of❓](https://hashnode.com/posts/what-an-api-consists-of/63d13d34107ed10d731538e8)
