decision-record-writer — independently scanned and version-tracked by SaferSkills.
SaferSkills independently audited decision-record-writer (Agent Skill) and scored it 100/100 (green). The audit ran 55 deterministic rules across Security, Supply Chain, Maintenance, Transparency, and Community; it found 0 high-severity and 0 lower-severity findings. The full rule-by-rule trace and per-finding evidence are below. Free, methodology-open.
Findings & checks · 0 flagged
Every scanned point with the score it earned and what moved between them.
First recorded scan — no prior version to compare against.
The primary manifest — the file an agent reads to learn what this artifact does.
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.
Accepted ADRs are never rewritten. When a decision is superseded:
This produces an immutable, versioned decision log that preserves rationale over time.
Before writing, determine:
If any of the above blocks progress, pause and ask.
Collect:
If the user cannot provide at least two plausible alternatives with tradeoffs, the ADR is not ready. Ask or decline.
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.
# {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{draft | proposed | accepted | deprecated | superseded}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>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.
For every chosen alternative, document at least:
Avoid listing only benefits. A decision without tradeoffs is not a decision — it is a fact.
The Negative Consequences section is mandatory. Every real decision has costs. If the user refuses to acknowledge negative consequences, document the refusal in Assumptions.
Every ADR must include at least one risk unless the decision is trivial. Risks without mitigations are still worth recording.
If you detect multiple distinct decisions in the request:
Do not bundle decisions even if the user insists. Offer to process them sequentially.
Refuse to write an ADR when:
When refusing, state the reason clearly and say what is needed to proceed.
Pause and ask when:
Ask only for the information that changes the outcome. Do not over-ask.
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
Before finalizing an ADR, verify:
If any mandatory item is missing, do not finalize.
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.
references/adr-format.md: full ADR template and field explanations~30 seconds. Free. No account. Every finding cites a rule and a line of evidence.