agentleFS
GustavoGutierrezGustavoGutierrez/engineering-skills/skills/decision-record-writer/SKILL.md

decision-record-writer

Skill0 starsChanged 5 months agoMITView on GitHub

Write Architecture Decision Records (ADRs) using append-only supersession model. Trigger: document architectural decision, create ADR, record decision rationale, architecture decision, technical decision log, or decision record.

1 other project uses a version of this.

Adopted by

Most recently taken up by

The file

Show its rules and text

We split this skill into the rules it gives an agent

20 rules. About 0 are common, 19 are distinctive, and 1 is very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈19

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 1

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Use this skill when the user asks to document an architectural or technical decision, create an ADR, record a decision rationale, or log a …

    line 26
  2. Distinctive:

    Write a new ADR with the same topic key

    line 32
  3. Distinctive:

    Is it one decision per ADR?: If the request bundles multiple independent decisions, recommend splitting into separate ADRs.

    line 45
  4. Distinctive:

    Can the decision be documented now?: If context, alternatives, or rationale are missing, ask for the missing pieces before writing.

    line 46
  5. Distinctive:

    If any of the above blocks progress, pause and ask.

    line 48
  6. Distinctive:

    If the user cannot provide at least two plausible alternatives with tradeoffs, the ADR is not ready.

    line 61
  7. Very specific:

    Use the required structure below.

    line 65
  8. Distinctive:

    Follow the ADR Quality Standards strictly.

    line 67
  9. Distinctive:

    Do not create a "straw man" alternative that no one would consider.

    line 134
  10. Distinctive:

    Avoid listing only benefits.

    line 142
  11. Distinctive:

    Every ADR must include at least one risk unless the decision is trivial.

    line 150
  12. Distinctive:

    If you detect multiple distinct decisions in the request:

    line 154
  13. Distinctive:

    Ask which one to address first

    line 158
  14. Distinctive:

    Do not bundle decisions even if the user insists.

    line 160
  15. Distinctive:

    Only one alternative is identified (no decision to make)

    line 165
  16. Distinctive:

    No tradeoffs are provided or obtainable

    line 166
  17. Distinctive:

    When refusing, state the reason clearly and say what is needed to proceed.

    line 170
  18. Distinctive:

    Ask only for the information that changes the outcome.

    line 180
  19. Distinctive:

    Before finalizing an ADR, verify:

    line 192
  20. Distinctive:

    If any mandatory item is missing, do not finalize.

    line 203

≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Decision Record Writer

When to Use

Use this skill when the user asks to document an architectural or technical decision, create an ADR, record a decision rationale, or log a technical choice. The output is a single Markdown ADR file following the append-only supersession model.

Core Principle: Append-Only / Supersession

Accepted ADRs are never rewritten. When a decision is superseded:

  • Write a new ADR with the same topic key
  • The new ADR links to the one it supersedes
  • Both remain in the record as an auditable trail

This produces an immutable, versioned decision log that preserves rationale over time.

Operating Workflow

Step 1. Evaluate the Request

Before writing, determine:

  1. Is there a real decision? Reject requests that ask for vague statements, non-decisions, or feature descriptions. A decision requires a choice between two or more plausible alternatives with tradeoffs.
  2. Is it one decision per ADR? If the request bundles multiple independent decisions, recommend splitting into separate ADRs.
  3. Can the decision be documented now? If context, alternatives, or rationale are missing, ask for the missing pieces before writing.

If any of the above blocks progress, pause and ask.

Step 2. Gather Minimal Viable Context

Collect:

  • What is the decision topic (one per ADR)?
  • What is the current state or problem?
  • What alternatives are on the table?
  • What tradeoffs exist between alternatives?
  • What consequences (positive and negative) follow from each alternative?
  • Who made or will make the decision and why now?

If the user cannot provide at least two plausible alternatives with tradeoffs, the ADR is not ready. Ask or decline.

Step 3. Write the ADR

Use the required structure below. Output as adr-{YYYY-MM-DD}-{slug}.md where slug is a short kebab-case topic descriptor.

Follow the ADR Quality Standards strictly.

Required ADR Structure

# {Decision Title}

## Status
{draft | proposed | accepted | deprecated | superseded}

## Date
YYYY-MM-DD

## Context
<What is the current situation or problem that requires a decision? Why is this decision necessary now?>

## Decision
<What was decided. Be specific about what will be done or what approach was chosen.>

## Alternatives Considered
- **{Alternative A}**: <description> — Tradeoff: <why this was not chosen or what tradeoffs it carries>
- **{Alternative B}**: <description> — Tradeoff: <...>

## Tradeoffs
<Explicit tradeoffs of the chosen path. Do not list only pros; document the costs and compromises.>

## Consequences

### Positive
- <intended positive outcome>

### Negative
- <unintended downside, risk, or cost. This section is mandatory.>

### Risks
- <identified risk>: <mitigation or fallback>

## Scope
<What is covered by this decision>
- <inclusion>

## Out of Scope
<What is explicitly NOT covered>
- <exclusion>

## Deciders
- <name or role>

## Assumptions
- <explicit assumption being made>

## Related ADRs
- Supersedes: <ADR-ID or "none">
- Superseded by: <ADR-ID or "none">
- Related: <ADR-ID or "none">

## Confidence
{high | medium | low} — <brief justification>

## Review Trigger
<Condition or date that should trigger re-evaluation of this decision>

ADR Quality Standards

Alternatives Must Be Plausible

Do not create a "straw man" alternative that no one would consider. Each alternative must be a real option that someone debated. If only one path exists, the decision is a statement, not a decision — ask the user to confirm.

Tradeoffs Must Be Explicit

For every chosen alternative, document at least:

  • One real cost or compromise
  • One reason the other alternatives were not chosen

Avoid listing only benefits. A decision without tradeoffs is not a decision — it is a fact.

Consequences Must Include Negatives

The Negative Consequences section is mandatory. Every real decision has costs. If the user refuses to acknowledge negative consequences, document the refusal in Assumptions.

Risks Must Be Identified

Every ADR must include at least one risk unless the decision is trivial. Risks without mitigations are still worth recording.

One Decision Per ADR

If you detect multiple distinct decisions in the request:

  1. Stop writing
  2. Tell the user which decisions are mixed together
  3. Recommend splitting into N separate ADRs
  4. Ask which one to address first

Do not bundle decisions even if the user insists. Offer to process them sequentially.

Refusal Conditions

Refuse to write an ADR when:

  • Only one alternative is identified (no decision to make)
  • No tradeoffs are provided or obtainable
  • The "decision" is a feature description in disguise
  • Critical context is missing and the user cannot provide it

When refusing, state the reason clearly and say what is needed to proceed.

When to Ask Before Writing

Pause and ask when:

  • The request mixes multiple decisions
  • The user cannot name at least two alternatives with tradeoffs
  • The decision is trivial (implementation detail with no architectural impact)
  • Context is incomplete but the user insists on proceeding

Ask only for the information that changes the outcome. Do not over-ask.

ADR Filename Convention

Derive from the decision topic. Use the date of the decision:

adr-2026-01-15-choose-postgres-for-primary-store.md

If draft: adr-2026-01-15-choose-postgres-for-primary-store-draft.md

Quality Checklist

Before finalizing an ADR, verify:

  1. Does the ADR contain at least two alternatives with tradeoffs?
  2. Does the Tradeoffs section include at least one cost of the chosen path?
  3. Does the Negative Consequences section include at least one real downside?
  4. Does the Risks section contain at least one risk (or explicitly state "no material risks identified")?
  5. Is the Status accurate?
  6. Is the Review Trigger defined?
  7. Is the Confidence rating justified?
  8. Are Related ADRs links complete?

If any mandatory item is missing, do not finalize.

Example Triggers

Good trigger: "We need to decide between Postgres and MongoDB for the primary store. The team is split and we have data volume of ~50k records/day."

→ Write an ADR with two real alternatives, tradeoffs, and consequences.

Bad trigger: "Create an ADR for our new auth system." → This is a feature, not a decision. Ask what architectural choice needs to be made.

Files to Read

  • references/adr-format.md: full ADR template and field explanations

References

  • ADR format: references/adr-format.md

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.