← Back to Skills Marketplace
Write Adr
by
Kevin Anderson
· GitHub ↗
· v1.1.2
· MIT-0
148
Downloads
0
Stars
1
Active Installs
2
Versions
Install in OpenClaw
/install write-adr
Description
Use when you want to generate Architecture Decision Records from this session. Triggers on "write ADRs", "document our decisions", "create decision records",...
Usage Guidance
This skill appears to do what it says: it scans your repo, extracts decisions from the session via subagents, asks you which to document, and writes MADR-formatted ADR markdown files. Before enabling or running it: 1) Confirm the repository contains the expected helper script (skills/adr-writing/scripts/next_adr_number.py) and the target directory (note the SKILL.md inconsistently references docs/adrs/ and docs/adr/), or it may fail or write files to an unexpected location. 2) Be aware subagents will read your repository files (commits, code) to collect context and will create files under docs/adr/ — review generated files before committing/pushing. 3) Verify you trust the referenced subskills (beagle-analysis:adr-decision-extraction and beagle-analysis:adr-writing) because this skill delegates model-driven work to them. 4) The skill does not request credentials or install external code, which reduces risk, but exercise the usual caution when allowing any skill to read and write your repository.
Capability Analysis
Type: OpenClaw Skill
Name: write-adr
Version: 1.1.2
The skill is a legitimate tool for automating the generation of Architecture Decision Records (ADRs). It uses standard read-only commands (git, ls, find) to gather context and orchestrates subagents to extract and write documentation to the local 'docs/adr/' directory, following a clear workflow with user confirmation steps (SKILL.md).
Capability Assessment
Purpose & Capability
The name/description matches the actions in SKILL.md: it scans repository context, extracts decisions, asks the user to confirm, and writes ADR files. Required binaries/credentials are none, which is coherent for a repo-local ADR generator.
Instruction Scope
Instructions intentionally read repository state (git branch, recent commits, ls/find of docs directories) and write ADR files under docs/adr/ (and refer to docs/adrs/ elsewhere). This file I/O is expected for the purpose, but there is an inconsistency in directory naming (docs/adrs/ vs docs/adr/) and a dependency on a repository-local script (skills/adr-writing/scripts/next_adr_number.py) that may not exist. The skill also asks subagents to "explore the codebase," which is within scope but means the subagents will read project files for context.
Install Mechanism
No install spec or external downloads — instruction-only — so nothing is written to disk by an installer and no third-party packages are pulled by the skill itself.
Credentials
The skill declares no required environment variables or credentials, which is appropriate. It does reference loading other skills (beagle-analysis:adr-decision-extraction and adr-writing); you should verify you trust those subskills.
Persistence & Privilege
The skill is not always-enabled and requests no elevated agent privileges. It sets disable-model-invocation: true for itself while instructing the agent to launch subagents that do model work — this is a design choice (not a direct risk), but it means the skill relies on other skills/subagents to perform model-driven steps. The skill will write files into the repository (docs/adr/), so expect persistent changes to project files.
How to Use
- Make sure OpenClaw is installed (local or Docker)
- Run the install command in chat:
/install write-adr - After installation, invoke the skill by name or use
/write-adr - Provide required inputs per the skill's parameter spec and get structured output
Version History
v1.1.2
**Expanded skill description, clarifies triggers and boundaries.**
- Skill description now details phrases that trigger execution and differentiates from related skills (adr-decision-extraction, adr-writing).
- Emphasizes orchestration of the end-to-end workflow: decision extraction, user confirmation, parallel ADR generation, and verification.
- Makes clear that this skill does not handle extraction alone or MADR template output.
- No workflow logic was changed—documentation only.
v1.1.1
**Improvements to ADR Generation Workflow**
- Enhanced decision extraction by analyzing session conversations with a dedicated subagent and presenting full details for user selection.
- Introduced user confirmation step: display comprehensive decision info and allow selective ADR creation (by number, "all", or "none").
- Refactored ADR number allocation: numbers are pre-assigned to decisions before subagents are launched in parallel, eliminating duplicate numbering.
- Standardized post-generation reporting and verification, including a checklist to ensure completeness and adherence to MADR format.
- Added instructions for ADR output location, initial template creation, and user follow-up for incomplete records.
Metadata
Frequently Asked Questions
What is Write Adr?
Use when you want to generate Architecture Decision Records from this session. Triggers on "write ADRs", "document our decisions", "create decision records",... It is an AI Agent Skill for Claude Code / OpenClaw, with 148 downloads so far.
How do I install Write Adr?
Run "/install write-adr" in the OpenClaw or Claude Code chat to install it in one step — no extra setup required.
Is Write Adr free?
Yes, Write Adr is completely free, licensed under MIT-0. You can download, install and use it at no cost.
Which platforms does Write Adr support?
Write Adr is cross-platform and runs anywhere OpenClaw / Claude Code is available (cross-platform).
Who created Write Adr?
It is built and maintained by Kevin Anderson (@anderskev); the current version is v1.1.2.
More Skills