@mrrlin-dev/mcp 0.2.5 → 0.3.0

package/dist/prompts/report-issue.md

@@ -105,31 +105,51 @@ short, ask **once**, and **skip any question the hint + log already answered**:
 If the user gives short or partial answers, accept them and move on. Never block
 the report on a perfect answer.
 
-### 4. Package the report (in ENGLISH)
+### 4. Package the report (in ENGLISH) — every string passes through `mrrlin-mcp redact`
 
 Translate the user's answers to English. Build a plain-text report. Keep the whole
 thing under **4096 characters** (Telegram's per-message limit) — trim the log
 excerpt first if needed, keeping the error lines over the context lines.
 
+**Hard rule:** every free-form string going into the report — the user's hint,
+their answers, the log excerpt — passes through the shipped scrubber first.
+There is no "but the log is already redacted by the logger" exception: the hint
+and the user's answers come from outside the logger, so they MUST be scrubbed
+here. Run each through:
+
+```bash
+HINT_REDACTED=$(printf %s "$USER_HINT_RAW" | mrrlin-mcp redact)
+EXCERPT_REDACTED=$(mrrlin-mcp redact < /tmp/excerpt-raw.txt)
+# (repeat for each user answer)
+```
+
+`mrrlin-mcp redact` reads stdin and writes the redacted bytes to stdout
+(empty input → empty output, exit 0). It uses the same regex set as the
+bridge logger (Bearer/JWT/GitHub-PAT/long-hex/long-base64). It is best-effort,
+not a guarantee — but it is the floor below which raw text is never allowed.
+
+Use the redacted versions to fill the template:
+
 ```
 🛠️ Mrrlin issue report
 When (UTC): <iso timestamp of the report>
 Mrrlin MCP: v<version if known>  |  OS: <platform>  |  Node: <version>
 
-▶ User hint (verbatim):
-<the text the operator passed with /report-issue, untranslated; "(none)" if absent>
+▶ User hint (verbatim, post-redaction):
+<HINT_REDACTED; "(none)" if absent>
 
 ▶ What the user was doing:
-<their answer, in English>
+<their answer, in English, post-redaction>
 
 ▶ Expected result:
-<their answer, or "(obvious from context: ...)", or "(not provided)">
+<their answer, post-redaction; or "(obvious from context: ...)"; or "(not provided)">
 
 ▶ Actual result / symptom:
-<their answer, in English>
+<their answer, in English, post-redaction>
 
-▶ Errors from bridge log (<filename>, matched by: <hint substring | sessionId | tail-fallback>):
-<the extracted error lines, verbatim — already redacted>
+▶ Errors from bridge log (<filename>, search strategy: <hint-driven | tail-fallback>):
+matched on: "<the exact line that triggered the cluster, truncated to 80 chars; or '(no hint provided)' for tail-fallback>"
+<EXCERPT_REDACTED — first the error lines, then the surrounding context>
 
 ▶ Context:
 session=<sessionId>  spanIds=<...>  window=<first ts>..<last ts>
@@ -180,10 +200,10 @@ why). Don't dump the raw report or the token back at them — just confirm.
 
 ## Rules
 
-- The operator's hint drives log search. Tail-only mode is the **fallback**, not the default.
-- Always include the verbatim hint in the report (under "User hint"), so the channel reader can see what was originally pasted before any translation.
+- **Every string going into the Telegram body comes out of `mrrlin-mcp redact`** — the user's hint, every translated user answer, the log excerpt. If your pipeline has a path that builds the body from raw text, you are doing it wrong. The bridge logger already redacts what it writes; the hint and the user's answers do NOT come from the logger and must be scrubbed here.
+- The operator's hint drives log search. Tail-fallback is the explicit fallback when no hint is provided.
+- `matched on:` must contain the literal line that triggered the cluster (truncated to 80 chars), so the channel reader can verify the match instead of trusting an LLM assertion.
 - Ask the user in their language; write the report in English. Skip any question the hint + log already answered.
 - Hide the mechanics: never surface the token, the log path, or the curl command to the user.
-- Forward log lines as-is — they're already secret-redacted. Do not paste anything that looks like a live token even if you see one.
 - One report = one `sendMessage` POST. Keep it under 4096 chars; use `sendDocument` only for the optional full log.
-- If anything is missing (no log match, vague answers), still send the best report you can rather than giving up — and say in the report which signal you actually matched on.
+- If anything is missing (no log match, vague answers), still send the best report you can rather than giving up — and say in the report which search strategy you used and what `matched on:` you found.

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@mrrlin-dev/mcp",
-  "version": "0.2.5",
+  "version": "0.3.0",
   "type": "module",
   "bin": {
     "mrrlin-mcp": "dist/bin.cjs"
@@ -20,12 +20,12 @@
     "@types/ws": "^8.18.1",
     "esbuild": "^0.24.0",
     "tsx": "^4.22.3",
-    "@mrrlin/client": "0.0.0",
     "@mrrlin/director-e2e": "0.0.0",
+    "@mrrlin/client": "0.0.0",
+    "@mrrlin/wiki": "0.0.0",
     "@mrrlin/codex-client": "0.0.0",
     "@mrrlin/schemas": "0.0.0",
-    "@mrrlin/tsconfig": "0.0.0",
-    "@mrrlin/wiki": "0.0.0"
+    "@mrrlin/tsconfig": "0.0.0"
   },
   "dependencies": {
     "@iarna/toml": "^2.2.5",