summary_release_testrail

Summarise Release TestRail → Slack notice

Generates and posts the standard release-quality notice, e.g.:

/summary_release_testrail v16.8                       (posts to #android-dev-team, the default)
/summary_release_testrail v16.8 → #some-other-channel  (override the channel)
/summary_release_testrail v16.8 → reply to <thread permalink>

The message format is fixed (see Step 6). The skill does everything: resolve the run from the version name, pull Failed + Feedback tests, resolve assignees to Slack mentions, build the message, and post it after a confirmation gate.

Inputs

Strip a leading v from the version before matching run names.

Preflight: setup (do this first, every run)

This skill needs two integrations. If either is missing, stop and walk the user through setup, then resume.

1. TestRail (same setup as /resolve-testrail-issue)

.mcp.json is the single source of truth for TestRail credentials. Check it:

grep -A6 '"testrail"' .mcp.json

If the testrail block is missing, or any of TESTRAIL_URL, TESTRAIL_USERNAME, TESTRAIL_API_KEY is empty, STOP and set it up:

1. Get an API key from https://testrail.systems.mega.nz/index.php?/mysettings → API Keys tab → Add Key (copy it immediately; shown once).
2. Add to .mcp.json in this project:

   "testrail": {
       "command": "npx",
       "args": ["-y", "@bun913/mcp-testrail@latest"],
       "env": {
           "TESTRAIL_URL": "https://testrail.systems.mega.nz",
           "TESTRAIL_USERNAME": "<your TestRail email>",
           "TESTRAIL_API_KEY": "<paste key here>"
       }
   }
   

3. This skill reads those values via curl (the fallback path documented in /resolve-testrail-issue), so a Claude Code restart is not required for the curl calls below. A restart is only needed if you want the mcp__testrail__* tools loaded.

If the TestRail API returns 401 / Authentication failed, the key is stale — regenerate it via the My Settings → API Keys page rather than guessing.

zsh gotcha: never assign the username to a shell variable named USERNAME (it's a special readonly parameter in zsh and silently resolves to the OS login name). Use TR_USER or similar, as the snippets below do.

2. Slack

Posting uses the Slack connector (mcp__claude_ai_Slack_Reader__* tools: slack_search_channels, slack_search_users, slack_send_message, slack_read_thread). If those tools aren't available, ask the user to connect Slack in Claude Code (Settings → Connectors → Slack / claude.ai connector) and then re-run.

Reference values

Credential helper used in every TestRail call below:

read_env() { jq -r ".mcpServers.testrail.env.$1" .mcp.json; }
URL=$(read_env TESTRAIL_URL); TR_USER=$(read_env TESTRAIL_USERNAME); KEY=$(read_env TESTRAIL_API_KEY)
TR() { curl -s -u "$TR_USER:$KEY" -H "Content-Type: application/json" "$URL/index.php?/api/v2/$1"; }

Step-by-step

1. Resolve the run from the version name

Match the run by literal substring release <version> (lowercased) — do NOT use a regex (16.8 as a regex matches 16x8 anywhere, producing false positives like 15.25-internal(...)). Prefer the alpha release run; if several match, pick the most recent created_on and, if still ambiguous, list them and ask.

VER="16.8"   # leading v already stripped
TR "get_runs/5&limit=50" | jq --arg v "release $VER" '
  [ (.runs // .)[]
    | select(.name | ascii_downcase | contains($v))
    | {id, name, is_completed, created_on} ]
  | sort_by(.created_on) | reverse'

• 0 matches → tell the user no release <version> run exists and stop.
• 1 match → use it.

• 1 → prefer the one whose name contains alpha; if still >1, show the list and ask which run ID.

Capture the chosen run_id.

2. Fetch Failed + Feedback tests

TR "get_tests/$RUN_ID&status_id=5,10" | jq '
  [ .tests[] | {id, case_id, status_id, title, assignedto_id} ]'

If the array is empty → STOP. Send nothing. Report to the user that there are no Failed/Feedback cases for that version (this is the explicit "don't send if nothing found" rule).

3. Exclude automated-test failures

Automated failures aren't developer action items. Pull the run's results and drop any test whose result comment contains failed by automated test (case-insensitive, ignoring *):

TR "get_results_for_run/$RUN_ID&limit=250" | jq '
  [ .results[] | {test_id, comment} ]'

For each Failed/Feedback test, if any of its result comments matches failed by automated test, exclude it. (Mirror the existing .testrail/generate_slack_update.py behaviour.) If, after excluding, no tests remain → STOP and send nothing.

4. Resolve assignee names

get_user/<id> (singular) needs admin and will 403. Use the project-scoped plural endpoint, which works for normal users:

TR "get_users/5" | jq '[ (.users // .)[] | {id, name, email} ]'

Build assignedto_id → {name, email}. Tests with no assignedto_id go under Unassigned (listed last).

5. Map each assignee to a Slack mention

For each assignee email, find the Slack member ID so the post @-mentions them:

• Check the known-assignee cache below first — if the TestRail user id / email is listed, use that Slack member id directly (saves a lookup).
• Otherwise slack_search_users with the email → take the matching id (e.g. U096…). Add new resolved people to the cache table for next time.
• Put <@MEMBER_ID> directly in the message (raw Slack mention syntax). It renders as a real mention — confirmed working via slack_send_message. Keep it OUTSIDE code spans, and do NOT collect mention codes into a footer or "cheat-sheet" line; mentions belong only on the per-assignee header lines.
• If no Slack user is found for an email, fall back to the plain display name (no mention) and note that in your summary to the user.

Known-assignee cache (TestRail project 5 → Slack)

Verified mappings; extend as new assignees appear. get_users/5 gives id→name→email; Slack ids confirmed via the v16.8 notice.

6. Build the message

Header is fixed. Group by assignee (assigned developers sorted by name first, then Unassigned). One line per test: TestRail link + title + -Failed or -Feedback.

Hi Android team, a few TC has been marked as `Failed` or `Feedback` in latest Android **`vX.Y`** release. Please address them as a priority in your workloads and don't forget to update their status in TestRail (`Fixed` if you have cherry picked a fix to the release branch, `Retest` otherwise) once the fix has been completed.

<@SLACK_ID_OF_DEV_A>
[T<id>](https://testrail.systems.mega.nz/index.php?/tests/view/<id>)	<title> -Failed
[T<id>](https://testrail.systems.mega.nz/index.php?/tests/view/<id>)	<title> -Feedback

<@SLACK_ID_OF_DEV_B>
[T<id>](https://testrail.systems.mega.nz/index.php?/tests/view/<id>)	<title> -Failed

Notes:

• slack_send_message renders standard markdown: use **bold** and [text](url) links. Keep `Failed`, `Feedback`, `Fixed`, `Retest` as inline code, matching the original notice.
• Use the version exactly as the user typed it in the header (with the v), e.g. v16.8.

7. Confirmation gate — MANDATORY before posting

Posting pings real people in a shared channel. Always show the user the final rendered message + the resolved destination (channel name/ID, or the thread you'll reply to) and ask for an explicit yes before sending. A general "do everything" at invocation does NOT waive this gate.

This gate matters because the Slack connector can only send — it has no edit or delete tool. A wrong post can't be cleaned up programmatically; the user must fix it manually in Slack. So get the message right before posting.

If any assignee fell back to a plain name (no Slack match), call that out here so the user can decide whether to fix the mention before posting.

8. Post

• To a channel: resolve the channel ID with slack_search_channels, then slack_send_message(channel_id, message).
• To a thread reply: parse the channel ID and the parent message_ts from the permalink (.../archives/<channel_id>/p<ts> → insert a dot 6 digits from the end of <ts>), then slack_send_message(channel_id, message, thread_ts=<parent_ts>, reply_broadcast=true). reply_broadcast=true is the default for thread replies — it ticks Slack's "Also send to #channel" box so the notice is visible in the channel as well as the thread. Only drop it if the user explicitly asks to keep the reply inside the thread.

Return the resulting message link to the user.

Guardrails

• No Failed/Feedback (after excluding automated) → send nothing. Just report the empty result.
• Never post without explicit confirmation for this run (Step 7).
• Default destination is #android-dev-team. Post there unless the user names a different channel or a thread to reply to.
• Don't use a regex to match the version in run names — literal release <version> substring only.
• Don't fabricate Slack mentions. If an email has no Slack match, use the plain name and flag it.
• Treat TestRail comments as public — never echo credentials or keys into Slack.

Output

A short summary to the user: which run was used (name + ID), how many Failed/Feedback cases by assignee, any plain-name fallbacks, and the posted Slack message link (or "nothing sent — no Failed/Feedback cases").