Authors the small, human-facing Decision Pause packet described by the Decision Pause Contract (in the framework bootstrap AGENT-BOOTSTRAP.md) without the agent hand-authoring the id, timestamps, or JSON. `create` mints a collision-free `D-<ULID>` id and writes the pending packet; `resolve` records the chosen option and rationale and moves it to resolved. A hand-picked sequential `D-{N}` is rejected, so two developers on parallel branches never collide (issue #272). The writer lives in `src/d...
Scanned 9/2/2026
Install to Claude Code
npx -y skills add Eliyce/paqad-ai --skill decision --agent claude-codeInstalls into .claude/skills of the current project.
Are you the author of Decision?
Add the live security badge to your README — it updates automatically with every re-scan.
[](https://www.skillsdirectory.com/skills/eliyce-decision)More formats (shields.io, HTML) on the badges page.
---
name: decision
description: Authors the small, human-facing Decision Pause packet described by the Decision Pause Contract (in the framework bootstrap AGENT-BOOTSTRAP.md) without the agent hand-authoring the id, timestamps, or JSON. `create` mints a collision-free `D-<ULID>` id and writes the pending packet; `resolve` records the chosen option and rationale and moves it to resolved. A hand-picked sequential `D-{N}` is rejected, so two developers on parallel branches never collide (issue #272). The writer lives in `src/decisions/authoring.ts`; this skill is the agent-side wrapper that invokes the bundled `create.mjs` / `resolve.mjs` scripts, the exact counterpart to `paqad-ai stage` on the stage-evidence ledger.
model_tier: fast
triggers:
- workflow:
- feature-development
- documentation-update
cacheable: false
cache_key_inputs: []
output_format: json
input_schema:
project_root:
type: path
required: false
description: Project root used to resolve `.paqad/decisions/`. Defaults to cwd.
category:
type: string
required: false
description: Decision Pause category (e.g. workflow-or-tool, architecture-path). Required by `create`.
title:
type: string
required: false
description: One-line question the decision answers. Required by `create`.
context:
type: string
required: false
description: The situation and trade-offs the user needs to weigh. Required by `create`.
options:
type: string[]
required: false
description: Two or more `<key>=<label>` option pairs. Required by `create`.
id:
type: string
required: false
description: The `D-<ULID>` id of a pending packet. Required by `resolve`.
chosen:
type: string
required: false
description: The `option_key` the user selected. Required by `resolve`.
---
## What It Does
Drives the pending → resolved lifecycle of a **Decision Pause packet** — the
readable decision record the Decision Pause Contract asks the agent to write to
`.paqad/decisions/`. The mechanical parts (minting the id, stamping timestamps,
writing and moving the JSON) run inside the bundled scripts so the agent supplies
only content, never id/lifecycle plumbing:
- `create.mjs` mints a collision-free `D-<ULID>` id and writes the pending packet
to `.paqad/decisions/pending/D-<ULID>.json`.
- `resolve.mjs` records the chosen option and rationale, moves the packet to
`.paqad/decisions/resolved/D-<ULID>.json`, and stamps `resolved_at`.
A hand-written sequential `D-{N}` is rejected: a new packet can only carry a
minted ULID id, so two developers on parallel branches never allocate the same id
(issue #184's collision-free form, now enforced at creation time — issue #272).
This is the _contract_ packet — distinct from the rich `DecisionPacket`
(`src/planning/decision-packet.ts`) that the automated intake / reuse pipeline
mints and consumes. See `references/decision-packet-contract.md`.
## Use This When
- A choice falls into a Decision Pause category and you must record it before
implementing (see the Decision Pause Contract's category list).
- You are about to hand-author a `D-{id}.json` packet — use this instead.
- The user has answered a pending decision and you need to resolve it and commit
the resolved packet with the change it justifies.
## Inputs
- `project_root` (path, optional) — defaults to cwd.
- For `create`: `category`, `title`, `context`, and two or more `<key>=<label>`
`options` (optional `recommendation` key).
- For `resolve`: the packet `id` and the `chosen` option key (optional rationale).
## Procedure
1. **Create.** Run:
```bash
node runtime/base/skills/decision/scripts/create.mjs <project-root> \
--category <category> --title <title> --context <context> \
--option <key>=<label> --option <key>=<label> [--recommendation <key>]
```
Capture the `id` from the JSON output.
2. **Present.** Surface the options to the user via the host's interactive UI (on
Claude Code, the `AskUserQuestion` tray). Ask one packet at a time in creation
order; ids sort chronologically.
3. **Resolve.** Once the user answers, run:
```bash
node runtime/base/skills/decision/scripts/resolve.mjs <project-root> <id> <chosen> [rationale]
```
4. **Commit.** Stage the resolved packet with the change it justifies, so a
reviewer and future `git blame` can see _why_ (delivery workflow).
## Output Contract
- `create.mjs` prints `{ "id", "path" }` (the minted id and the pending file).
- `resolve.mjs` prints `{ "path" }` (the resolved file).
- Both exit non-zero with a message on stderr for a usage or validation error
(missing field, fewer than two options, non-ULID id, unknown `chosen` key).
## Escalate / Stop Conditions
- **Never hand-edit** the packet JSON, its timestamps, or the pending/resolved
move — always go through the scripts.
- If `create` reports a validation error, fix the content and re-run; do not work
around it by writing the file yourself.
- If the interactive UI is unavailable (non-interactive run, hook context), stop
and wait until the resolved packet exists, per the contract's fallback.
## Resources
- `references/decision-packet-contract.md` — packet shape, lifecycle, and rules.
Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.
No comments yet. Be the first to comment!