Apply brand tone and writing style fixes

docs/middleware/budget-guard.mdx

@@ -243,6 +243,6 @@ The cost estimate before the call is a character-based approximation unless you
 
 Pricing changes faster than this package releases. Confirm the built-in rates against your provider's own pricing page and override the ones that matter to you.
 
-The session and daily budgets are tracked in the cache, not the database, and reset on their own window rather than at a fixed clock boundary. They require a key resolver that can identify the session or the user. Without one, the middleware silently skips those budgets.
+The middleware tracks the session and daily budgets in the cache, not the database. Each budget resets on its own window rather than at a fixed clock boundary. They require a key resolver that can identify the session or the user. Without one, the middleware silently skips those budgets.
 
 When a run's model changes between steps, the per-request budget prices the earlier steps' accrued usage at the current step's rate, which is an approximation.

docs/middleware/injection-guard.mdx

@@ -448,7 +448,7 @@ Intercept logs detections under a distinct message, `Prompt injection attempt de
 ]
 ```
 
-The `degraded_from` key appears only when the configured action could not be applied. The `field` value is a dot path into the edited arguments, so nested values are reported precisely.
+The `degraded_from` key appears only when the configured action could not be applied. The `field` value is a dot path into the edited arguments, so it pinpoints nested values precisely.
 
 ### Custom callbacks on a resumed run
 

docs/middleware/pii-redactor.mdx

@@ -335,7 +335,7 @@ This ignores other supported entity types.
 
 ### Allowing specific email addresses
 
-Use `allowedEmails` when known safe email addresses should not be redacted or blocked.
+Use `allowedEmails` when you want the middleware to ignore specific known safe email addresses.
 
 ```php
 public function middleware(): array
@@ -522,7 +522,7 @@ The middleware logs detections found in approval decisions under a distinct mess
 
 The `degraded_from` key appears only when the run continued despite a configured action that could not be applied. A blocked run never reports it.
 
-The `field` value is a dot path into the edited arguments, so nested values are reported precisely, for example `arguments.filters.contact.email`.
+The `field` value is a dot path into the edited arguments, so it pinpoints nested values precisely, for example `arguments.filters.contact.email`.
 
 ### Disabling approval decision scanning
 
@@ -548,7 +548,7 @@ public function middleware(): array
 ```
 
 <Warning>
-    Detection offsets on this path are relative to the decision value the detection was found in, not to the prompt. Offsets exist to drive redaction, which cannot apply to a resumed run. Custom callbacks should therefore use the detection `type` and `value` rather than `start` and `length`.
+    Detection offsets on this path are relative to the decision value that contains the detection, not to the prompt. Offsets exist to drive redaction, which cannot apply to a resumed run. Custom callbacks should therefore use the detection `type` and `value` rather than `start` and `length`.
 </Warning>
 
 ## Production rollout

docs/middleware/tool-approval-guard.mdx

@@ -113,7 +113,7 @@ Entities listed in `block_entities` stop the run regardless of the configured ac
 
 A proposed argument that matches an injection pattern suggests that content the middleware never saw manipulated the model. It uses the same patterns as [Injection Guard](/middleware/injection-guard).
 
-**This scan is off by default.** Prose a model writes for a human reader routinely contains phrases the patterns match, such as "You are now subscribed to weekly updates.", "From now on we will email you every Monday.", "System: scheduled maintenance at 3pm." None of those indicate manipulation in that context.
+**This scan is off by default.** Prose a model writes for a human reader routinely contains phrases the patterns match. Examples include "You are now subscribed to weekly updates.", "From now on we will email you every Monday.", "System: scheduled maintenance at 3pm." None of those indicate manipulation in that context.
 
 Enable it when a proposed argument feeds **another model or agent** rather than a person: