mgiovani/ai-arsenal / skills/docs-adr
docs-adr
VerifiedCreate a numbered Architecture Decision Record (ADR) documenting a technical decision with context, alternatives, and consequences. Trigger on "create an ADR", "document this architectural decision", "record why we chose X", or "write an ADR for [decision]". Not for a decision still open for discussion (use docs-rfc, RFCs propose, ADRs record) and not for bootstrapping a project's whole docs/ tree from scratch (use docs-init).
Create Architecture Decision Record
Create a new Architecture Decision Record (ADR) documenting an architectural decision.
Anti-hallucination guidelines
ADRs document real decisions about real code: every claim in the ADR must be verifiable in the repo, not assumed. Before writing:
- Verify the technology exists: if the ADR mentions "Redis", confirm Redis is actually used somewhere in the codebase.
- Reference actual files: grep/glob to find real file paths, never invent one.
- Quote real code: if citing a pattern, find an actual example of it.
- Check current state: the Context section must reflect verified reality, not a plausible-sounding guess.
Workflow
Phase 1: Parse Arguments
- Extract the decision title from the command arguments. If no title was given, stop here and ask the user for one (and optionally which variant): don't invent a placeholder title or proceed to the later phases.
- Check for a variant keyword as the leading token:
lightweight,full, ornygard. A matching word inside the title itself (e.g. "Full-Text Search") is not a variant keyword, only strip it when it's a standalone token preceding the title. - If a variant keyword is found, strip it from the title.
- Default variant:
nygard.
Phase 2: Determine ADR Number
- Scan
docs/adr/for files matchingXXXX-*. - Find the highest existing number and increment by 1 (start at
0001if none exist). - Format as a 4-digit zero-padded number (e.g.
0001,0023).
Phase 3: Sanitize Title for Filename
Convert the title to kebab-case, lowercase, special characters stripped.
Example: "Use Redis for Caching" -> use-redis-for-caching.
Phase 4: Gather Context
Use the Task tool with the Explore agent, when available, to search the codebase for the decision topic: current implementation (if any), related config files, dependencies involved, and existing documentation. Ask it to return verified file paths and relevant snippets, not summaries it can't back up.
If the Task tool isn't available, run the equivalent searches directly instead:
e.g. grep -rn "<topic>" across source files, find . -name "*.config.*" or
docker-compose.yml for infra-flavored decisions, find . -name "*schema*" -o -name "*models*" for data-layer decisions. Either path, only include context you actually
found; an ADR with no verifiable context is a red flag, not something to pad with
plausible-sounding filler.
Phase 5: Load and Populate Template
- Templates live in
assets/templates/:nygard.md(default),lightweight.md,full.md. - Load the selected template and grep it for every
{{TOKEN}}placeholder it actually contains, the three templates use different token sets (e.g. nygard has{{CONTEXT}}; lightweight has{{PROBLEM}},{{DECISION}},{{ALTERNATIVES}},{{CONSEQUENCES}},{{NOTES}}; full has a longer set including{{AUTHORS}},{{STAKEHOLDERS}},{{OPTION_1_NAME}}, etc.). Don't assume a fixed list: fill whatever the loaded template actually contains. {{ADR_NUMBER}},{{ADR_TITLE}}, and{{DATE}}(YYYY-MM-DD) appear in all three; fill those from Phases 1-3 regardless of variant.- After substitution, scan the rendered output for any leftover
{{...}}: zero unresolved tokens before writing the file.
Phase 6: Create ADR File
- Filename:
docs/adr/XXXX-kebab-case-title.md. - Create
docs/adr/if it doesn't exist. - Write the populated content with initial Status set to "Proposed".
Phase 7: Report Creation
Report the ADR number, title, and file path, plus next steps (e.g. review with the team, flip Status to Accepted once approved).
Template Variants
| Variant | Sections | Use when |
|---|---|---|
| nygard (default) | Status, Context, Decision, Consequences | Most decisions, balanced detail |
| lightweight | Status, Decision, Rationale | Simple, straightforward decisions |
| full | Status, Context, Decision Drivers, Considered Options, Decision, Consequences (Positive/Negative/Neutral), Pros and Cons, Related Decisions, References | Complex, high-impact decisions |
Usage Examples
docs-adr "Database Migration Strategy"
-> docs/adr/0004-database-migration-strategy.md (nygard, next available number).
docs-adr lightweight "Use Redis for Session Storage"
-> strips "lightweight", verifies Redis is actually referenced in the repo, writes
docs/adr/0005-use-redis-for-session-storage.md with just Status/Decision/Rationale.
docs-adr full "Adopt Event-Driven Architecture"
-> writes the full variant with Considered Options and split Consequences, citing real messaging/event code found during context gathering.
ADR Numbering & Status Lifecycle
- First ADR is conventionally
0001-record-architecture-decisions.md(meta-ADR); subsequent ones auto-increment. - Status progresses: Proposed (default on creation) -> Accepted -> Deprecated or Superseded (by a later ADR, which should link back to this one).
Notes
- One decision per ADR: split unrelated decisions into separate records.
- Write in imperative language ("we will", not "we should").
- Document the real reasons a decision was made, including trade-offs and downsides, not the idealized version.
- Cross-link: reference related or superseded ADRs by number.
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| rea-changelog-updatemorluto/reaPrepare or rewrite REA release changelogs and GitHub release notes from pinned Git history, with verified contributor thanks and Release Please synchronization. | Stars: 469k | Updated | Claude CodeCodex | Verified | Install |
| academy-guideanthropics/skillsStop and check this skill before finishing any reply to a question about how to use Claude or a Claude product — it recommends matching courses, tutorials, and use cases from Claude Academy (academy.claude.com), Anthropic's learning hub. Trigger on: "how do I", "how can I", "getting started with", "what can Claude do", "teach me", "learn to use"; questions about artifacts, projects, skills, plugins, connectors, MCP; requests about rolling Claude out to a team, class, or organization; and any ask for training materials, onboarding content, or learning resources. Use it when the user is learning how to use a feature or product — not when they are mid-task and just want the task done. This skill composes with other skills: after consulting product documentation to answer how a Claude feature works, also check here for a matching course or tutorial — a docs-grounded answer and an Academy recommendation belong together. Only recommend on a strong match; never invent Academy content. | Stars: 197k | Updated | Claude CodeCodex | Verified | Install |
| claude-apianthropics/skillsReference for the Claude API / Anthropic SDK — model ids, pricing, params, streaming, tool use, MCP, agents, caching, token counting, model migration. TRIGGER — read BEFORE opening the target file; don't skip because it "looks like a one-liner" — whenever: the prompt names Claude/Anthropic in any form (Claude, Anthropic, Fable, Opus, Sonnet, Haiku, `anthropic`, `@anthropic-ai`, `claude-*`, `us.anthropic.*`, `[1m]`); the user asks about an LLM (pricing/model choice/limits/caching) — never answer from memory; OR the task is LLM-shaped with provider unstated (agent/MCP/tool-definition/multi-agent/RAG/LLM-judge/computer-use; generate/summarize/extract/classify/rewrite/converse over NL; debugging refusals/cutoffs/streaming/tool-calls/tokens). SKIP only when another provider is being worked on (overrides all triggers): OpenAI/GPT/Gemini/Llama/Mistral/Cohere/Ollama named in the query; OR `grep -rE 'openai|langchain_openai|google.generativeai|genai|mistralai|cohere|ollama'` over the project hits (run this grep FIRST if no provider named — don't Read the file). | Stars: 197k | Updated | Claude CodeCodex | Unverified | Install |
| doc-coauthoringanthropics/skillsGuide users through a structured workflow for co-authoring documentation. Use when user wants to write documentation, proposals, technical specs, decision docs, or similar structured content. This workflow helps users efficiently transfer context, refine content through iteration, and verify the doc works for readers. Trigger when user mentions writing docs, creating proposals, drafting specs, or similar documentation tasks. | Stars: 197k | Updated | Claude CodeCodex | Verified | Install |