# Pickup Task

> Direct agent to execute a tracked task/plan from the vault, updating its status before and after the work.

- **Type:** Skill
- **Install:** `agentstack add skill-chicken-noodle-chris-tome-pickup-task`
- **Verified:** Yes — security-reviewed for prompt injection and unsafe behavior
- **Seller:** [chicken-noodle-chris](https://agentstack.voostack.com/s/chicken-noodle-chris)
- **Installs:** 0
- **Category:** [Security](https://agentstack.voostack.com/c/security)
- **Latest version:** 0.1.0
- **License:** MIT
- **Upstream author:** [chicken-noodle-chris](https://github.com/chicken-noodle-chris)
- **Source:** https://github.com/chicken-noodle-chris/tome/tree/main/skills/pickup-task

## Install

```sh
agentstack add skill-chicken-noodle-chris-tome-pickup-task
```

Requires the [AgentStack CLI](https://agentstack.voostack.com/docs/cli). Works with Claude Code, Cursor, and any MCP-compatible agent.

## About

Optional input: which task or plan to pick up (a task ID, a plan name, or a description).

This skill executes work that's already been planned. Use `scripts/tome.py` (`tome help`)
to manage status changes — don't hand-execute git or status moves. `tome` is on PATH in
Bash; if not found, fall back to `python "$TOME_PLUGIN_ROOT/scripts/tome.py" `. It
resolves which vault via `--vault` / walking up from cwd / `VAULT_ROOT`; bare paths like
`wiki/SCHEMA.md` are relative to the vault root, not the plugin root.

**The user is always happy to answer questions.** If intent or scope is unclear — which task,
how far to take it, an ambiguity in the plan — ask.

1. **Prep on Haiku (gate).** Steps 2–3 are mechanical — a git pull, priming, and a task
   lookup — and don't need a strong model. If you are **not** running as Haiku, stop and
   ask whether to switch to Haiku for the prep (the user may decline and stay on their
   current model — either is fine; you can't switch your own model, only they can).
   Already Haiku, or once they've answered: continue. The tier that matters for the
   actual *work* is a separate decision, made at step 4 once the task is known.

2. **Prime yourself on the vault.** Run `tome sync` to pull, then `tome prime --full`
   (skip if already primed this session) — prints SCHEMA.md and the index in one shot.

3. **Locate the task and its plan.** If the user named a task or plan, find it. If they
   named a **milestone** instead (an id like `m-0` or its title — the epic layer, see
   `wiki/SCHEMA.md`), resume it rather than picking one task in isolation: run `tome task
   task list --milestone  --plain` to list its open children (backlog.md already
   orders each status group by ordinal), then pick the next one — an In Progress child
   first (someone left it mid-flight), else the highest-priority To Do child, ties broken
   by list order. Confirm the pick with the user if it's not clear-cut. Selection data
   comes entirely from that list output; there's no separate tome command for this. With
   no name given at all, check the board (`tome task task list --plain`) and the project's
   live plans (`plans/`, not `plans/archive/`) and confirm which one they mean.

4. **Check the agent-tier label.** With the task located, read its `agent:` label
   (tier: `haiku` ` — accepts either, resolving
   the other half if linked (a plan without a task, or a task without a plan, is normal).
   Sets the plan `active`, moves the task to In Progress (`-a @me`), logs `work-started`,
   syncs, then prints the task text and the full plan body as your working context — read
   it in full, follow its `[[wikilinks]]`, and if you have concerns stop and discuss with
   the user before continuing.

6. **Do the task work.** Execute the plan in the relevant code repo (not the vault),
   following that repo's `CLAUDE.md`. Stick to the plan's scope; if you hit a fork it
   doesn't resolve, ask rather than guess. Verify per the plan's verification section and
   report results honestly.

7. **Commit, then present the work.** Commit by default. Show the user what changed and the
   verification results. The code repo is separate from the vault and follows its own
   commit conventions — check its `CLAUDE.md`. Push only after the user approves.

8. **Close out the tracking.** Once the work has landed: `tome done  --summary "..."`
   citing the shipping commit (`--as superseded`/`--as abandoned` instead of the default
   `done`, if that's how it landed). This archives the plan (moves it to `plans/archive/`,
   regenerates its project hub and the index), checks the task's acceptance criteria,
   closes and completes it with the summary, re-points its `--ref` at the archived path,
   logs `done`, and syncs — one command, no deferred sweep. **Umbrella plans:** if the
   task is one phase of a milestone plan shared by sibling phase tasks, `tome done
   ` closes only that task and leaves the plan active (it prints how many open
   siblings kept it alive); the plan archives automatically when the last sibling closes.
   Don't try to close the shared plan slug while phases remain open — it's refused unless
   `--force`.

## Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

- **Author:** [chicken-noodle-chris](https://github.com/chicken-noodle-chris)
- **Source:** [chicken-noodle-chris/tome](https://github.com/chicken-noodle-chris/tome)
- **License:** MIT

Install and usage instructions live in the source repository linked above.

## Pricing

- **Free** — Free

## Security capabilities

Automated source analysis of v0.1.0 — what this tool can access:

- **Network access:** no
- **Filesystem access:** no
- **Shell / process execution:** no
- **Environment & secrets:** no
- **Dynamic code execution:** no

*"Yes" means the capability is present in the source — more access means more to trust, not that it is unsafe.*


## Versions

- **0.1.0** — security scan: passed — Imported from the upstream source.

## Links

- Listing page: https://agentstack.voostack.com/l/skill-chicken-noodle-chris-tome-pickup-task
- Seller: https://agentstack.voostack.com/s/chicken-noodle-chris
- Browse the marketplace: https://agentstack.voostack.com/browse

---
Listed on AgentStack — the marketplace for AI agent skills and MCP servers. Every listing is security-reviewed. Creators keep 70%.