Secret protection should begin before a prompt is assembled. If Nerva collects an entire repository and tries to remove API keys afterward, it has already made the sensitive-data problem harder than necessary. The planned context firewall should minimize first, detect second, redact third, and ask before any external request.

What should be treated as sensitive

Common examples include .env files, private keys, bearer tokens, database URLs, cloud credentials, cookies, and copied production logs. Path rules help, but content detection matters too because a secret can be stored in an ordinary file such as config.ts.

Redaction should not pretend to be perfect. A scanner may miss a custom token format or flag a harmless string. The user should see uncertain matches and choose whether to keep the request local, edit the context, or cancel.

Removed from request:
- .env.local
- Authorization header in logs, line 42
- DATABASE_URL value in config.ts

Destination: local model
Cloud request: not made

Do not leak secrets through tool output

A command can expose a secret even when the input files were safe. Nerva should apply output filtering before showing or sending terminal results, while preserving enough surrounding information to debug the failure. It should also avoid writing raw tool output into long-term task history by default.

For a backend developer, this design makes AI assistance more practical: code can be inspected and explained while credentials remain outside the model context. Redaction is not a promise of perfect security, but it is a concrete layer that users can review.

The useful design question

The useful question here is not whether an AI agent can produce an impressive demo. It is whether the workflow stays understandable when the task is incomplete, the repository is messy, or the model is wrong. For this part of Nerva, the focus is treating secrets as a data-flow problem. That means each state should have a visible owner, a small input, and a result that another part of the system can inspect.

A concrete scenario

Consider blocking .env files and redacting tokens before a request leaves the local context boundary. A good Nerva run would first state the scope, identify the files or operation it needs, and show the next proposed step. It would keep the original request separate from model-generated instructions. If an action needs approval, the user would see what will happen before it happens. Afterward, the task history would retain the command, result, and verification status instead of reducing everything to a confident paragraph.

This approach also makes a failure useful. A denied permission, unavailable provider, missing package, or changed file is information about the task state. Nerva should preserve it and explain the smallest recovery option. It should not silently broaden the request, retry a side effect, or claim completion because a model supplied plausible-looking text.

What I would test first

For an early implementation, I would test three paths: a normal successful run, a denied action, and an interrupted run resumed after the workspace changes. The key test is whether redaction is visible and whether outputs can leak a secret back. In each path, the interface should make clear what Nerva knows, what it proposed, what actually ran, and what remains unverified.

Useful acceptance checks include:

  • the user can inspect the selected context before model execution;
  • sensitive paths and credentials are excluded or redacted;
  • every side effect has a narrow permission decision;
  • failures remain attached to the step that caused them;
  • a final answer links to evidence instead of implying that unrun work succeeded.

Nerva is still an upcoming project, so these are design and implementation goals rather than claims that a finished product already provides them. That distinction matters: a trustworthy agent should earn confidence through observable behavior, not through a feature list.

The boundary that matters in production

Redaction needs adversarial tests. Include tokens split across lines, values copied into command output, URLs containing credentials, and secrets produced by a tool rather than read from a file. A redactor should report that it removed content without revealing the original value. It should also fail closed when it cannot determine whether a high-risk value is safe to send.

For the first vertical slice, keep the boundary small enough to explain in a code review. Define the input, the decision, the side effect, and the evidence for completion. Add a test for the denied path as well as the successful path. This is where a local-first design earns trust: the user can inspect the context and history on the device, see when a provider is involved, and stop before an action crosses the permission boundary. The design should remain honest about its limits. Nerva is planned software, not a claim that every workflow can be made safe automatically. A clear “cannot verify” state is more useful than a polished but unsupported success message.

A review of this design should use a small fixture repository and record both allowed and rejected outcomes. That keeps the discussion concrete: reviewers can compare the requested scope with the selected context, action, and evidence. It also prevents Nerva from being judged only by how persuasive its response sounds.

The design is informed by the boundaries I have had to think through in CryptoEx and SentinelFi: backend work becomes easier to review when context, permissions, side effects, and verification are explicit.