@agentsignalnet/claude-plugin 0.2.68 → 0.2.70

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.
@@ -1,7 +1,7 @@
1
1
  {
2
2
  "name": "agentsignal",
3
3
  "displayName": "AgentSignal",
4
- "version": "0.2.68",
4
+ "version": "0.2.70",
5
5
  "description": "Page a human from inside Claude Code, or become an addressable agent that other agents can reach.",
6
6
  "author": {
7
7
  "name": "Data Group AI",
@@ -38,13 +38,28 @@ that is not hypothetical, it is what happened. An agent already registered as
38
38
  `claude mcp add`, and ended up with two servers whose tools collide by name and
39
39
  which would each have spent a use of the same single-use token.
40
40
 
41
+ THE COMMAND COMES FROM THE PLUGIN, THE TOOLS COME FROM THE SERVER, AND THEY ARE
42
+ INSTALLED SEPARATELY. This is the thing to get right, because an earlier version
43
+ of this file got it wrong in a way that reads as correct: it told somebody with
44
+ nothing to run `claude mcp add`, then told them to run `/agentsignal:join`. That
45
+ command can never appear from `claude mcp add` -- a server contributes TOOLS,
46
+ never slash commands -- so the recipient did as instructed, found no command,
47
+ and reported the install as broken when it had worked. The token went unspent
48
+ and somebody spent an evening on it.
49
+
50
+ So there are three states, not two, and the invitation has to name which one the
51
+ reader is in rather than offering a choice:
52
+
41
53
  Print this, and nothing that competes with it:
42
54
 
43
55
  > Run `/agentsignal:join <the token>`.
44
56
  >
45
- > If that command does not exist and you have no AgentSignal tools at all, you
46
- > need the server first — see below. If you DO have AgentSignal tools, the join
47
- > command is the whole of it: do not install anything.
57
+ > **If that command does not exist**, check whether you have AgentSignal tools
58
+ > at all — something like `join_channel` or `notify`.
59
+ >
60
+ > - **Tools but no command**: you have the server without the plugin. Nothing to
61
+ > install. Ask your Claude to call `join_channel` with the token above.
62
+ > - **Neither**: install the plugin — see below. It brings the server with it.
48
63
 
49
64
  The join path is not merely shorter. It is the one that keeps the machine on a
50
65
  single server: `join_channel` adds a channel to the server already running, and
@@ -52,26 +67,40 @@ the tools for it turn on in that same session — the SDK sends
52
67
  `tools/list_changed` and the client picks them up. Nothing is installed, no
53
68
  file is edited, and there is no restart, because there is no new process.
54
69
 
55
- **Only for somebody with no AgentSignal at all**, put the install below the
56
- join line, clearly subordinate to it:
70
+ **Only for somebody with no AgentSignal at all**, put the install below the join
71
+ line, clearly subordinate to it. THE PLUGIN, not the bare server, because it is
72
+ what carries `/agentsignal:join` -- and it bundles the server, so this is one
73
+ install rather than two:
74
+
75
+ ```
76
+ claude plugin marketplace add https://agentsignal.net/marketplace.json
77
+ claude plugin install agentsignal@agentsignal
78
+ ```
79
+
80
+ Then restart Claude, and then the join command. A new PROCESS needs a restart; a
81
+ new channel does not, so say which of the two they are doing — otherwise the
82
+ restart looks mandatory to somebody who does not need it.
83
+
84
+ If they would rather not install a plugin, the server alone can be added:
57
85
 
58
86
  ```
59
87
  claude mcp add agentsignal -e AGENTSIGNAL_JOIN_TOKEN=<the token> -- npx -y @agentsignalnet/mcp
60
88
  ```
61
89
 
62
- If `claude mcp add` refuses with "MCP server agentsignal already exists in
63
- local config", that is not a naming problem to work around — it is the answer
64
- to the test above. They already have one. Tell them to stop and run
90
+ SAY, IN THE SAME BREATH, THAT THIS ROUTE ENDS WITH NO SLASH COMMAND. Their
91
+ Claude redeems the token itself: a session that starts holding an unredeemed one
92
+ is told, before it sees a single tool, that its first action is `join_channel`.
93
+ Without that sentence they will restart, look for `/agentsignal:join`, and
94
+ conclude the install failed — which is precisely what happened.
95
+
96
+ If `claude mcp add` refuses with "MCP server agentsignal already exists in local
97
+ config", that is not a naming problem to work around — it is the answer to the
98
+ test above. They already have one. Tell them to stop and run
65
99
  `/agentsignal:join` instead of renaming. An earlier version of this file
66
- suggested `agentsignal-net` to dodge that collision, which turned the one
67
- signal that a server already existed into a step to route around.
68
-
69
- Under the install, and only there: restart Claude. A new PROCESS needs one; a
70
- new channel does not. Say which they are doing so the restart does not look
71
- mandatory to somebody who does not need it. The server introduces itself — a
72
- Claude that starts holding an unredeemed join token is told, before it sees a
73
- single tool, that its first action is `join_channel` — so do not paste a script
74
- for them to run.
100
+ suggested `agentsignal-net` to dodge that collision, which turned the one signal
101
+ that a server already existed into a step to route around.
102
+
103
+ Do not paste a script for them to run.
75
104
 
76
105
  Say when the token expires, and say that it is single-use if it is.
77
106
 
package/package.json CHANGED
@@ -1,6 +1,6 @@
1
1
  {
2
2
  "name": "@agentsignalnet/claude-plugin",
3
- "version": "0.2.68",
3
+ "version": "0.2.70",
4
4
  "publishConfig": {
5
5
  "access": "public"
6
6
  },
@@ -1920,12 +1920,47 @@ const PAGE = `<!doctype html>
1920
1920
  }
1921
1921
  }
1922
1922
 
1923
+ /*
1924
+ Clear the refusal as soon as it stops being true. Left up, "Choose who this
1925
+ goes to." is still on screen after they have chosen, which reads as the
1926
+ choice not having registered.
1927
+ */
1928
+ for (const id of ['to', 'say']) {
1929
+ document.getElementById(id).addEventListener('input', () => {
1930
+ const note = document.getElementById('sent');
1931
+ if (note.textContent.startsWith('Choose') || note.textContent.startsWith('Type')) {
1932
+ note.textContent = '';
1933
+ }
1934
+ });
1935
+ }
1936
+
1923
1937
  document.getElementById('compose').onsubmit = async (event) => {
1924
1938
  event.preventDefault();
1925
1939
  const to = document.getElementById('to').value;
1926
1940
  const say = document.getElementById('say');
1927
1941
  const note = document.getElementById('sent');
1928
- if (!to || !say.value.trim()) return;
1942
+
1943
+ /*
1944
+ SAY WHY, rather than doing nothing.
1945
+
1946
+ This was one bare early return covering both cases, so clicking Send with no
1947
+ recipient chosen -- the state the form STARTS in -- did nothing at all:
1948
+ no error, no hint, the typed message still sitting there. Reported as
1949
+ "the send button is not working", which is exactly what it looks like
1950
+ from the outside, and the button was working perfectly.
1951
+
1952
+ Not a disabled button instead. A disabled Send is the same silent
1953
+ failure wearing a different hat: it is clicked, nothing happens, and
1954
+ nothing explains which of the two fields is the problem.
1955
+ */
1956
+ if (!to) {
1957
+ note.textContent = 'Choose who this goes to.';
1958
+ return;
1959
+ }
1960
+ if (!say.value.trim()) {
1961
+ note.textContent = 'Type something to send.';
1962
+ return;
1963
+ }
1929
1964
  note.textContent = 'sending…';
1930
1965
  // The channel on screen, said explicitly. A message composed under a
1931
1966
  // Sandbox header and delivered into a different channel is the worst