draft

A command that turns a decision already made in the current plan into one new architecture decision record (ADR).

In plain words
What is it for?
Use it to identify a lasting choice in a plan and write its context, decision, alternatives, and consequences into a proposed record.
Why use it?
Plans are often replaced or forgotten, while the technical choices inside them may still matter to future work.

Command

Install

Getting it into your agent

One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.

agentmods
npx agentmods add commands/mbeacom/adrkit/draft
Clone the repo
git clone --depth 1 https://github.com/mbeacom/adrkit
Per session 16 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 579 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. Scan, not verified.
Origin original No closer match found in the catalogue.
Token cost

What it costs to keep this loaded

Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.

ModelPer sessionOnce invoked
Fable 5 $0.00016 $0.00579
Opus 5 $0.00008 $0.00290
Sonnet 5 $0.00003 $0.00116
Haiku 4.5 $0.00002 $0.00058

Measured yesterday against content hash cafacb9882a5, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

draft scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured yesterday.

A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.

Nothing flagged

None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.

packages/adapters/spec-kit/commands/draft.md · 53 lines

How it starts

The opening of the file, as written. The whole thing — 53 lines — stays where its author put it; the contents beside it link to each section on GitHub.

adrkit — draft a decision from this plan

Turns a decision the plan already made into a record that outlives it. Plans get superseded and forgotten; the decisions inside them are what the next feature needs and never gets.

User Input

$ARGUMENTS

Steps

  1. Read the current plan. Identify the decision worth recording — a choice between real alternatives with consequences that outlast this feature. If the plan made no such choice, say so and stop. Not every plan contains an ADR, and manufacturing one is worse than not writing it.
  2. Run {SCRIPT} with the decision's title as arguments. Write the title as the decision itself, in the imperative — "Adopt Postgres for the read model", not "Database choice".
  3. The script prints { "id", "path" }. It has scaffolded a schema-valid draft at the next free id; it has not written the reasoning.
  4. Open that file and fill it in from the plan:
    • Context — the forces that made this a decision. Pull them from the plan; do not invent constraints the plan never stated.
    • Decision — what was chosen, in the plan's own terms.
    • Options considered — the alternatives the plan actually weighed, with why each lost. If the plan weighed none, record that honestly rather than inventing a strawman to reject.
    • Consequences — what gets easier, what gets harder, and what would make this worth revisiting.
    • Frontmatter — affects patterns for the paths this governs, relatesTo for records it touches, and a review.tier matching the blast radius.
  5. Leave the status as draft. Ratification is a human act, and this command does not perform it.
  6. Run /speckit.adrkit.check afterwards to confirm the new record is valid and does not conflict with an existing decision.

Notes

  • This is the only command in the extension that writes, and it is deliberately not reachable from any hook. A plan-phase hook that created records unprompted would manufacture decision memory instead of recording it.
  • It writes exactly one new file and never edits an existing record. If the target path is already taken it refuses rather than overwriting (exit 1).
  • A plan must exist. Drafting "from the plan artifact" without a plan is not a degraded mode, it is a different and worse thing.

Read the full file on GitHub · 53 lines

Changes

What this file has done since we first saw it

Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.

  1. yesterday First seen · 53 lines · 16 tokens per session scan A cafacb9882a5

Subscribe to this mod's changes

draft is a command published in the GitHub repository mbeacom/adrkit (11 stars, last pushed 2d ago), licensed Apache-2.0. It adds 16 tokens to every session and 579 once invoked, about $0.0001 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.