add-to-queue
/add-to-queue
Add saved-up or new work to the work queue (scripts/.queue, `ravenseat`) as tracks of the right size (K 90-120 min, D about 40, D+ 25-30, U about 60; small non-blocking tasks in the batch queue), with risk class, project and feature, ownership, dependencies and owner questions. Also takes a plan entry from the plan register (`/add-to-queue PL-…`) and sets the plan field. Use when the user says "add to the queue", "/add-to-queue …", or wants to send work to the continuous queue. Builds nothing itself.
- Author
- vegardjorg ↗
- Repository
- ravenseat ↗
- Licence
- MIT
- Last repository change
- 13.11.2026
Assessment and security
- Files in this package
- 1
- Scripts
- 0
- HTML comments
- 0
- URLs in SKILL.md
- 0
No signals detected. Repository changed on 13.11.2026, without references to retired models or APIs.
MIT, open: use and modify while retaining author attribution.
What it does and when to use it
Add saved-up or new work to the work queue (scripts/.queue, `ravenseat`) as tracks of the right size (K 90-120 min, D about 40, D+ 25-30, U about 60; small non-blocking tasks in the batch queue), with risk class, project and feature, ownership, dependencies and owner questions. Also takes a plan entry from the plan register (`/add-to-queue PL-…`) and sets the plan field. Use when the user says "add to the queue", "/add-to-queue …", or wants to send work to the continuous queue. Builds nothing itself.
Triggers from the description
Original author description
Add saved-up or new work to the work queue (scripts/.queue, `ravenseat`) as tracks of the right size (K 90-120 min, D about 40, D+ 25-30, U about 60; small non-blocking tasks in the batch queue), with risk class, project and feature, ownership, dependencies and owner questions. Also takes a plan entry from the plan register (`/add-to-queue PL-…`) and sets the plan field. Use when the user says "add to the queue", "/add-to-queue …", or wants to send work to the continuous queue. Builds nothing itself.
How it works
SKILL.md has no level-two headings · 75 lines, 944 words
SKILL.md preview
---
name: add-to-queue
description: Add saved-up or new work to the work queue (scripts/.queue, `ravenseat`) as tracks of the right size (K 90-120 min, D about 40, D+ 25-30, U about 60; small non-blocking tasks in the batch queue), with risk class, project and feature, ownership, dependencies and owner questions. Also takes a plan entry from the plan register (`/add-to-queue PL-…`) and sets the plan field. Use when the user says "add to the queue", "/add-to-queue …", or wants to send work to the continuous queue. Builds nothing itself.
---
# Add to the work queue
What goes in: **$ARGUMENTS** (if empty, ask what to add, or use the tasks the user described in the conversation).
You only ADD work. Do not build, do not start agents, do not run `ravenseat start`/`set`; the orchestrator does that.
Background: `${CLAUDE_PLUGIN_ROOT}/skills/parallel-wave/SKILL.md` and `docs/queue/queue.md`.
**Lite mode:** run `ravenseat mode` first. If it says `lite`, skip steps 0 to 5: add one task per thing the user asked for
with `ravenseat add "<short title>" --mission "<self-contained mission: what, why, where, how to check it>"` (add
`--files <paths>` when you know the files it owns, `--depends <ids>` when it must wait for other tasks, `--priority 1-10`
only when the user gave a priority). No class,
size, value, project, feature, splitting or packing; ids, class K and the defaults come by themselves, and nothing
is refused for being small. A task that needs the owner's decision first still waits on a registered question: use
`ravenseat add --json '{"tasks":[{"title":"…","mission":"…","waiting_on":["Q-1"]}],"question":{"Q-1":{"type":"question","text":"…","recommendation":"…"}}}'`.
Then do step 7.
0. **Plan entry as input.** If the argument is a plan entry (`PL-…`), or the work belongs to one
(`ravenseat plan show --search "…"`), fetch it with `ravenseat plan get <id> --json` and check:
maturity `ready` (otherwise say what is missing and suggest `/plans <id>` first), no open blocking owner questions
(otherwise tasks that must wait get them in `waiting_on`), and that content only found in gitignored notes is lifted
into the mission (worktree agents cannot see gitignored files). Every task gets `"plan": "<entry-id>"` and `project`
from the entry. `add` refuses an unknown entry, and the entry becomes `in-queue` by itself.
1. **Split into tracks of the right size per class** (fixed overhead is a large share of a small track):
- **K** (client/UI/ordinary code), weight 1.0: **90-120 min of content**. No K under about 30 min alone; pack it
with related K work in the same project/feature. Split only if the track would pass about 250k context.
- **D** (database/RPC/contract changes), weight 1.3: **about 40 min** (30-50), at most 3 migrations.
- **D+** (new tables, retention/anonymisation, unauthenticated endpoints, patching a shared function), weight 1.8:
**25-30 min**, one function family. Never merged with other work.
- **U** (investigation), weight 0.6: about 60 min.
- Keep apart: K and D; tasks with different owner questions; unrelated hotspot files. Parallelism before cohesion.
- **Small-task guard (`add` enforces it):** 10 min or less: not in the queue; ask the owner "fix now
(recommended) or batch queue?". Over 10 min but under the lower target for the class and value 7 or less:
batch queue (`"batched": true`). Unsure value: ask with a recommended value. Deliberate own track:
`"own_track": "<reason>"`. Not for D+, for tasks others depend on, or for tasks with open owner questions.
`content_min` = the time the content would take as a K track (estimate = content_min × weight).
2. **Write one self-contained track description per track** to `scripts/.queue/mission/<id>.md`: what and why, "Read
also" (paths), numbered content, tests (red first), review needs and "Do NOT touch".
3. **Map ownership with Grep/Glob, do not guess:** `owned_files` (exact paths, or folders ending in "/"), `owner_db`
(functions created or redefined), `shared_patch` (shared functions patched), `migr_tens` (0 without migrations).
Then `reads`: the files the task builds on but does not change. Grep for the imports and calls of what the task
uses (the modules it imports, the functions and types it calls, the config it reads) and list those files, or a
folder with "/". A file in `owned_files` is never also in `reads`. Leave out what everything uses (a shared
utility nobody changes): `reads` makes the task wait while another task changes the file.
4. **Project and feature (REQUIRED):** every task gets `project` (from `ravenseat projects`) and `feature` (a short
name the owner recognises). Tasks with visible UI get `"labels": ["visual"]`. Unsure of the project:
`ravenseat classify --file <in.json>`. No fitting project: create one first:
`ravenseat project <pid> --title "<clear title>" --prefix <PREFIX>- [--plan docs/<plan>.md] [--area <oid>]`.
4b. **Cloud or local:** set `requires` (list) when the task needs something a cloud session lacks: `db`, `web`,
`local-files`, `unit`. Class D/D+ and DB ownership keep a task local anyway.
5. **Make one JSON file** (e.g. `scripts/.queue/mission/<prefix>-in.json`):
```json
{ "tasks": [{ "id": "EX-1a", "title": "short title", "domain": "web",
"project": "example", "feature": "Login page", "labels": ["visual"],
"class": "K", "mission": "scripts/.queue/mission/EX-1a.md", "value": 7, "content_min": 90,
"migr_tens": 0, "depends_on": [], "waiting_on": ["Q-EX-1"], "owned_files": ["src/login/"], "reads": ["src/api/session.ts"],
"owner_db": [], "shared_patch": [], "requires": [], "plan": "PL-EX-1", "source": "docs/plan.md § 3",
"added_by": "<session>" }],
"question": { "Q-EX-1": { "type": "question", "text": "The question, clear for the owner",
"recommendation": "…", "source": "docs/plan.md § 3" } } }
```
`id`: unique, letters/digits/"."/"_"/"-". `waiting_on` takes ONLY registered question ids (with text and type,
from before or in the same file); `add` and `edit` refuse anything else, and the whole file stops. Types:
`question`, `condition`, `lawyer`, `owner-task`. When the owner answers (`ravenseat resolve <id> --answer "…"`), every
waiting task gets the answer appended to its mission.
5b. **Acceptance criteria:** a task with user-visible behaviour gets 1-3 short, checkable criteria
(`"acceptance": ["..."]` in the file, or `ravenseat add ... --accept "..."`); they are locked once the task starts.
6. **Add and check:** `ravenseat add --file <file.json>` and `ravenseat show --project <pid>`.
`add` rejects the whole file on error: fix and retry.
7. **Report briefly:** the ids added, project and feature, dependencies, and owner questions (with recommendation).
No commit needed: `scripts/.queue/` is gitignored.
8. **Investigations and proposals become plan entries:** if it is an idea or proposal not ready to build, register it
with `ravenseat plan new --file <json>` (`"priority": "owner"`, `"maturity": "idea"`, `"priority_set_by": "agent"`, a
reason, and a linked owner question). Mention the entry id in the report.
Versions
The database has no other skills named add-to-queue.
Database snapshot: 13.11.2026 · package version 8847b1974dde (SHA-256 digest of the entire package)
Downloads in the last 30 days: 0 · ranking
Report a problem with this skill: harmful content, incorrect descriptions or an author’s removal request.
Similar skills
- 90pick
Pick a task from the work queue (`ravenseat`) and do it yourself in this session, following the same protocol as a queue track. Use when the user types "/pick" (suggestions for what can be picked now) or "/pick <id>" (take exactly that task), or says "pick a task", "what can I… - 100setup
Set up the current project for the work queue (Ravenseat plugin): queue.config.json, docs, CLAUDE.queue.md, .gitignore, recommended settings, the empty queue, a smoke test, and optionally the user''s own TODOs and plans as queue entries. Use when the user types… - 100new-plan
Queue a UNIT OF WORK in the plan/todo queue of a documentation-led repo, tracing to an existing ADR — names the owning ADR(s), scope, exit criteria mapped to acceptance criteria, dependencies, and queue position. Use when the user says "add a plan item", "queue this work",… - 95ideas-and-queue
Keep listening for work from the review page, take approved items from the queue and build them, and answer requests for ideas. Use after any review round, whenever you would otherwise stop, or when told to take the next item from the Markback queue. - 100cloud-tasks-queue-setup
Configure cloud tasks queue setup operations. Auto-activating skill for GCP Skills. Triggers on: cloud tasks queue setup, cloud tasks queue setup Part of the GCP Skills skill category. Use when working with cloud tasks queue setup functionality. Trigger with phrases like "cloud… - 90add-entry
Add or correct a Gearlist catalog entry. Use when asked to add a tool, MCP server, skill or platform to this repository, or to promote a row from stars-queue.json.