fix(plugin): preserve original MCP failure evidence / 补齐失败现场与采集缺口

claude/.claude-plugin/plugin.json

@@ -1,7 +1,7 @@
 {
   "name": "chatcut",
   "displayName": "ChatCut",
-  "version": "1.10.21",
+  "version": "1.10.22",
   "description": "ChatCut plugin for editable video editing and creation workflows from Claude Code.",
   "author": {
     "name": "ChatCut",

claude/skills/chatcut-plugin-basics-claude/SKILL.md

@@ -30,6 +30,8 @@ When an existing connection or OAuth attempt reports an error, or tools remain a
 
 Also use connection diagnostics when previously available tools disappear or the host reports an unfamiliar MCP/protocol error. Capture the current-host snapshot described in that reference; preserve unknown errors and stages.
 
+For a failed hosted ChatCut tool request after connection, immediately preserve the original tool error and any available failure time, conversation ID and local call ID in the dedicated failure record described there. The Agent gathers these facts; do not ask the user to manually reconstruct them. Startup-status notices and later successful checks do not establish what happened to that tool request.
+
 1. **Server not registered:** Offer to register the Hosted MCP server directly at user scope. Do not install, reinstall, repair, or enable a plugin to obtain the server:
 
 ```sh

claude/skills/chatcut-plugin-basics-claude/references/connection-diagnostics.md

@@ -2,7 +2,7 @@
 
 Use this branch when a registered hosted ChatCut connection fails, including
 OAuth failing before a browser opens, or authentication completing without usable
-tools. Preserve evidence before reconnecting or restarting. This version gathers
+tools, and failures of already available hosted tools. Preserve evidence before reconnecting or restarting. This version gathers
 evidence; it does not repair configuration, credentials, network settings, or the
 original connection.
 
@@ -104,6 +104,49 @@ Only include ChatCut tool names; keep tool arguments and results out.
 
 ## Run once for the reported failure
 
+### Preserve the failed call before independent checks
+
+When the host exposes the original failed tool result or runtime event, the Agent
+copies only its error and available identifiers into a new dedicated
+`<name>.mcp-failure.json`. Do this before retries, reconnects or later probes.
+Do not read a conversation transcript or collect tool arguments/results to build
+it. Use the host's existing error metadata and supported read-only runtime logs;
+the user does not fill this record. Keep the full original error message, code,
+cause and stack when available, including unfamiliar failures.
+
+```json
+{
+  "schemaVersion": 1,
+  "source": "host_tool_result",
+  "serverName": "chatcut",
+  "occurredAt": null,
+  "conversationId": null,
+  "callId": null,
+  "operation": "list_projects",
+  "requestId": null,
+  "error": { "message": "Exact original error; add original code/cause/stack when available" }
+}
+```
+
+Supported sources are `host_tool_result`, `host_runtime_event`, and
+`captured_stderr`. `occurredAt` is the failed event's actual timestamp with
+timezone, when exposed; never substitute the current time, file modification
+time, snapshot capture or later probe time. Omit unknown optional IDs or leave
+them null. A local `callId` and `conversationId` identify client-side work and
+are useful even if the request never reaches ChatCut; they are not server
+request IDs. Copy a `requestId` only when the original failed call actually
+provides it. All supplied facts remain labelled as Agent-supplied evidence.
+
+Prefer `--failure-record <name.mcp-failure.json>` to `--original-error` when
+this structured source is available; choose one input. The collector centers
+its bounded log window on the supplied failure time and uses its conversation
+ID when available. A missing time remains an explicit gap. For an exec caller,
+use already retained stderr; discarded stderr cannot be recovered by this Skill.
+Do not start another authenticated session to manufacture replacement evidence.
+
+If only the original error text is accessible, use the text-file fallback below.
+Do not call a report with only startup-status notices a captured tool failure.
+
 1. Obtain the existing server's registered URL through the host. Preserve its
    environment. The helper accepts ChatCut production/beta/dev endpoints and
    loopback development servers; it reports unsupported endpoints rather than
@@ -153,6 +196,17 @@ records from three 128 KiB tails. Missing/rotated/denied logs remain explicit
 gaps. The known Windows MXC `LOCALAPPDATA` sandbox redirect falls back to the
 real user's `AppData/Local` without changing permissions. An already supplied alternative App log root can be selected with
 `--codex-app-log-dir`; do not scan browser profiles or session transcripts.
+
+The report names selected/omitted native files and byte ranges that were not read.
+For an existing reported gap, use `--codex-app-log-file <known codex-desktop-*.log>`
+to select at most three known native files. With exactly one selected file,
+`--codex-app-log-offset <nonnegative byte offset>` reads at most 128 KiB from
+that segment; unread ranges remain explicit. Preserve the original failure's
+known time/window. Do not repeatedly upload replacement reports or scan entire
+log history to fill an unknown failure time. App startup notices cover startup;
+they do not cover every later tool-call failure. SQLite/native log extraction
+retains available local conversation/call IDs without promoting them to original
+server request IDs or independently confirmed attempt associations.
 For a separately chosen future reproduction, Codex's explicit `log_dir` setting
 can enable its native plaintext log. Do not change global settings or relaunch
 the user's host automatically. Existing logs cannot be reconstructed retroactively.