[bmestrallet]

Why

The sales pipeline skills flag a deal as stale when its card has no comment for 30 days. The compact task lists did not carry the date of a task's last comment, so an assistant read the comments of every open card, one call each (EXO-91997).

How

• list_tasks, list_assigned_tasks and list_unscheduled_tasks: each compact item carries last_comment_date and last_comment_author (username, display_name) of its latest comment, a reply or not; both are absent for a task without comments.
• comments_count counts the replies, in the compact items and in get_task_by_id. The change log of a full task (include_change_log, list_project_activity_since) has one commented entry per comment and per reply.
• CommentService#getCommentsSummaries(Collection<Long>) (through CommentStorage and CommentHandler) reads a page's comments in two statements, whatever the number of tasks: a GROUP BY task for the count and the latest time, then one lookup of the comments at those times for the author, the highest id on a tie. It replaces the per-task count of the compact items. No subquery per comment row, no schema change.
• The three tool descriptions name the new fields.
• About 146 more characters on a commented item (date 48, author 98), none on an uncommented one.
• Tests: TestCommentDAO#testGetCommentsSummaries runs both statements on HSQLDB (replies counted, a reply as the latest, a tie, a same-time comment of another task, absent tasks); TaskMcpToolTest pins one summaries call and no per-task count for a page of 100 tasks, the latest comment's date and author, and the change log's replies. Each pin was mutation-verified. The query plans were not measured on a real database.

Knowledge: Meeds-io/eng-standards#672, Meeds-io/eng-standards#670

This change is classified N2 at 7ab377738 (a DAO query on a list path, changed definitions of three existing MCP tools; no schema, entity, ACL or new tool) — its approver must be an Archi/Dev who knows it is N2, not an approval on AI review alone.

🤖 Generated with Claude Code

[sonarqubecloud]

Quality Gate passed

Issues
0 New issues
0 Accepted issues

Measures
0 Security Hotspots
83.3% Coverage on New Code
0.0% Duplication on New Code

See analysis details on SonarQube Cloud

[bmestrallet]

Self-review close-out — head 7ab377738

#	Source	Finding	Status
1	delivery checks §1	get_task_by_id's count with replies, its change log built from top-level comments only	✅ fixed 4fb1bf913, pinned by theChangeLogOfAFullTaskHoldsTheReplies
2	reviewer 🟡	a max() subquery correlated per comment row on TASK_COMMENTS (no TASK_ID index on PostgreSQL)	✅ fixed 7ab377738: a grouped query and an author lookup, no subquery
3	reviewer 🟡	the last_comment_date pin ran under the constant date-format stub	✅ fixed 7ab377738: each date formatted as its instant; creation-date and due-date swaps fail it
4	reviewer 🟢	an unclear TaskListItemModel Javadoc sentence	✅ fixed 7ab377738

The closing round, by a reviewer that had not seen the fixes, returned nothing to fix.

Verified conform: one summaries call per page for the three compact-item tools and no per-task count (pinned on 100 tasks); replies counted and a reply as the latest comment (HSQLDB); the tie and the same-time comment of another task kept apart; both fields absent without comments (NON_EMPTY); layering DAO → storage → service → tool; additive interface methods, no other implementor; /** */ on every new method.

Tests: mvn -o verify -pl services -Pcoverage → Tests run: 378, Failures: 0, Errors: 0, all coverage checks met, at 7ab377738. Mutation-verified pins: tie-break inverted, same-time check removed, count without replies, per-task count restored, latest comment dropped, latest date swapped for the creation or due date, full count top-level only, replies dropped from the change log. Query plans not measured on a real database.

Classification: N2 at 7ab377738 — a DAO query on a list path, changed definitions of three existing MCP tools; no schema, entity, ACL or new tool. Its approver must be an Archi/Dev who knows it is N2, not an approval on AI review alone.

[bmestrallet]

Local rig, through the MCP — task-management-services.jar built at 7ab377738 on the feature/ai-contribution rig (HSQLDB), task 208 with one comment and one reply:

• list_assigned_tasks / list_tasks: comments_count: 2, last_comment_date = the reply's (23:39:11.271), last_comment_author {root, Root Root}; the uncommented tasks carry neither field.
• get_task_by_id: comments_count: 2.
• list_tasks include_change_log: two commented entries, the comment and the reply.
• list_tasks limit 100: 100 items, no task, comment, MCP or SQL error in platform.log.