Agents that call your API: how are you authenticating and scoping them safely?

Parker Le ⭐14 · Mar 2, 2026 14:44
Excited about tool use, scared about prompt injection turning into arbitrary HTTP. What patterns are people shipping beyond 'please be good model'?
15 replies
Parker Ahmed ⭐175 · Mar 2, 2026 16:44
Separate OAuth client per agent workflow with least-privilege scopes — never reuse human tokens.
Emerson Carter ⭐48 · Mar 2, 2026 20:44
We proxy tool calls through a service that validates arguments against a JSON schema before hitting internal APIs.
Finley Tan ⭐191 · Mar 3, 2026 00:44
Allow-lists of endpoints per environment stopped the accidental production delete scenario we nightmare about.
Quinn Carter ⭐143 · Mar 3, 2026 04:44
Human confirmation step for irreversible actions — agents propose, humans click, we log both IDs.
Quinn Walker ⭐171 · Mar 3, 2026 08:44
Rate limits per agent session caught a loop bug that would have hammered payments.
Drew Khan ⭐138 · Mar 3, 2026 12:44
We strip HTML from retrieved web pages before they enter the context window to reduce injection surface.
Hayden Le ⭐116 · Mar 3, 2026 16:44
Mutations require idempotency keys generated server-side, not by the model.
Parker Bennett ⭐153 · Mar 3, 2026 20:44
Audit log includes full tool payload hash so security can replay without storing sensitive arguments in plaintext forever.
Casey Pham ⭐38 · Mar 4, 2026 00:44
Network policies isolate the agent runtime from databases it should never touch, even if code regresses.
Quinn Tan ⭐20 · Mar 4, 2026 04:44
Red-team exercises with dedicated budget found three creative ways to exfiltrate secrets — patched before external beta.
Casey Hoang ⭐30 · Mar 4, 2026 08:44
We do not let models compose SQL — only choose from parameterised templates maintained by engineers.
Reese Hoang ⭐68 · Mar 4, 2026 12:44
Short-lived JWTs for tool calls reduced blast radius when a session was compromised in testing.
Parker Walker ⭐73 · Mar 4, 2026 16:44
Teaching support to recognise when an attacker is trying to jailbreak through the chat saved a real account once.
Emerson Nguyen ⭐112 · Mar 4, 2026 20:44
Static analysis on prompt templates catches forgotten string concatenation that used to slip through review.
Jordan Nguyen ⭐34 · Mar 5, 2026 00:44
Principle: treat the model as an untrusted user — same gates you would put on a public form.

Join the conversation.

Log in to reply