Check Duplicates Skill

Canonical deduplication gate for tusk. Run this before inserting any task to avoid creating redundant work.

Usage

code

/check-dupes check "<summary>" [--domain <domain>] [--threshold <float>] [--json]
/check-dupes scan [--domain <domain>] [--status <status>] [--threshold <float>] [--json]
/check-dupes similar <id> [--domain <domain>] [--threshold <float>] [--json]

Commands

`check "<summary>"`

Pre-insert gate. Checks if a summary is a duplicate of any open task.

bash

tusk dupes check "<summary>" --domain <domain>

When to use: Before every INSERT INTO tasks — in /next-task, /analysis-to-tasks, or any manual task creation.

`scan`

Find all duplicate pairs among open tasks.

bash

tusk dupes scan --status "To Do"

When to use: During /groom-backlog to surface duplicates for cleanup.

`similar <id>`

Find tasks related to a given task ID (uses a lower threshold of 0.6).

bash

tusk dupes similar <id>

When to use: Exploratory — when investigating whether a task overlaps with other work.

Interpreting Results

Exit Codes

Code	Meaning
0	No duplicates found — safe to create
1	Duplicates found — skip or review
2	Error (task not found, DB error)

Thresholds

Score	Interpretation	Action
>= 0.82	Duplicate	Skip the INSERT
0.60 - 0.82	Partial overlap	Present to user for decision
< 0.60	New	Safe to create

JSON Output

Use --json for programmatic consumption:

json

{
  "duplicates": [
    {"id": 42, "summary": "...", "domain": "Backend", "similarity": 0.95}
  ]
}

Integration Points

This skill is referenced by:

•/next-task — Before creating deferred tasks
•/groom-backlog — Scanning for duplicate pairs