smooth-operator-mcp 3.0.1 → 3.0.2
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- package/dist/smooth-operator.mjs +72 -36
- package/dist/smooth-operator.mjs.map +1 -1
- package/docs/mcp-server.md +12 -6
- package/package.json +1 -1
package/docs/mcp-server.md
CHANGED
|
@@ -147,7 +147,9 @@ smooth-operator
|
|
|
147
147
|
```
|
|
148
148
|
|
|
149
149
|
Remote mode is rejected unless the token is at least 32 characters. Do not
|
|
150
|
-
use a token from a shell history, checked-in file, or shared log.
|
|
150
|
+
use a token from a shell history, checked-in file, or shared log. POST requests
|
|
151
|
+
must declare `Content-Type: application/json`; unsupported media types are
|
|
152
|
+
rejected before their body is read. A reverse
|
|
151
153
|
proxy can add TLS and network access controls, but it does not replace the
|
|
152
154
|
application token, Host/Origin allowlists, or request-size limit. Allowed host
|
|
153
155
|
and origin values are hostnames without a scheme; browser preflight requests
|
|
@@ -409,13 +411,16 @@ disabled explicitly with `SMOOTH_OPERATOR_ALLOW_EVAL=false`.
|
|
|
409
411
|
`browser_evaluate` is page JavaScript and is available by default (set
|
|
410
412
|
`SMOOTH_OPERATOR_ALLOW_EVAL=false` when it is not wanted). `browser_exec`
|
|
411
413
|
accepts only a JSON array of validated browser actions; it is not a shell,
|
|
412
|
-
Python, or arbitrary code runner.
|
|
414
|
+
Python, or arbitrary code runner. Its explicit `evaluate` action still follows
|
|
415
|
+
the page-evaluation policy. Destructive batch actions require explicit
|
|
413
416
|
confirmation. `browser_wait_for_human` pauses for an operator to complete a
|
|
414
417
|
visible sign-in or challenge. `browser_solve_challenge` is an internal
|
|
415
418
|
connected-AI observe/act/verify loop: it returns bounded evidence and is
|
|
416
419
|
successful only when a fresh final classification explicitly reports the
|
|
417
|
-
challenge absent
|
|
418
|
-
|
|
420
|
+
challenge absent; present and exhausted cycles report the remaining attempt
|
|
421
|
+
budget and `automation_exhausted` is the final non-success state.
|
|
422
|
+
`browser_close_session`
|
|
423
|
+
closes the one native browser session by its explicit session identifier.
|
|
419
424
|
|
|
420
425
|
Actions that leave a usable page—navigation, click, input, select, scroll, key,
|
|
421
426
|
back, forward, and reload—accept optional `includeSnapshot: true`. The action
|
|
@@ -470,8 +475,9 @@ The server publishes read-only resources:
|
|
|
470
475
|
Resource output is bounded and follows the same redaction and policy rules as
|
|
471
476
|
tool output. The capabilities resource also reports the native defaults and
|
|
472
477
|
effective feature flags for local browser tools, page evaluation, stealth, and
|
|
473
|
-
behavioral timing
|
|
474
|
-
|
|
478
|
+
behavioral timing. Its challenge metadata includes the default and maximum
|
|
479
|
+
connected-AI attempt budgets and states that success requires an explicit
|
|
480
|
+
absent classification.
|
|
475
481
|
|
|
476
482
|
### Prompts
|
|
477
483
|
|