adr
Records a lightweight Architecture Decision Record for a decision made in this session.
Skill by Tonours
Install
Works with Claude Code, Claude apps, Codex, Cursor, Gemini CLI, GitHub Copilot, VS Code, OpenCode, Goose, Windsurf. Read the repository before you give any tool access to your data.
npx agentmarket install skill/tonours-adr
Asks which of your tools to add it to (Claude Code, Codex, Cursor, VS Code, Windsurf, Gemini CLI, Claude Desktop) and shows the change first. See https://agentmart.dev/cli.
Get the weekly diff
The weekly diff: one email a week with new agent tools, notable updates and grade changes, plus security alerts for tools you follow. Double opt-in, one-click unsubscribe.
About
Reviews
Write a reviewNo reviews yet. Installed it? Reviews come from verified installers only. How reviews work
Permissions
What this skill can do and what it asks for, from its SKILL.md. Checked Oct 25, 2026.
- Tools
- No tools found
- Secrets
- None asked for
- Install scripts
- None found
- Runs as
- Node.js package
Network hosts it talks to 1
codeload.github.com· found in install command
Files it touches
~/.claude/skills/adr
What's inside
- Skill name
adr- Description
- Records a lightweight Architecture Decision Record for a decision made in this session. Use when the user invokes /adr or asks to capture an architectural decision, trade-off, supersession, or why a non-obvious technical choice was made.
- License
- MIT
- Instructions
- 92 lines in SKILL.md · view on GitHub
Before installing
- Ships executable scripts
Files (5)
- ADR-FORMAT.md
- SKILL.md
- VALIDATOR-PORTABILITY.md · scripts/
- adr-validation.mjs · scripts/
- apply-adr.mjs · scripts/
SKILL.md (first lines)
# /adr — record an Architecture Decision Record
Capture a decision made in this session as an immutable ADR. Write ADR content
in English, regardless of the conversation language. Use
`scripts/apply-adr.mjs` for deterministic file writes, numbering, supersession
metadata, the `docs/adr/README.md` index, and the `CLAUDE.md` pointer. `ADR-FORMAT.md` is the reference for exact
template and decision rules; read it when the decision needs those details.
You run inside the current session, so you already see the conversation. Do
NOT read any transcript file — synthesize the decision from what was discussed
plus the working diff.
## Procedure
1. **Identify the decision.** From the conversation (and `$ARGUMENTS` if given),
state in one sentence what was decided. Ground it with token-light git
evidence first: `git status --short`, `git diff --name-only`, and
`git diff --stat`. Read targeted diffs only for files needed to understand
the decision; avoid dumping a large full diff.
2. **Inventory existing ADRs.** If `docs/adr/` exists, read its file list,
frontmatter, and titles. Read the latest 3 ADRs by number, expanding to 5
only when recent history looks relevant. Search existing
ADR titles, `tags`, `affected_components`, and body text for terms from the
new decision, then read the full text of any likely matches. If
`scripts/validate-adrs` exists, run `node scripts/validate-adrs` before
drafting; stop and report any existing ADR integrity failure before adding a
new ADR. Do not repair pre-existing ADR or index failures while
recording a new decision; those must be fixed as a separate maintenance task.
This is the anti-hallucination gate: never claim an ADR is superseded unless
you can cite a local `docs/adr/NNNN-*.md` file and explain the relationship.
If no grounded candidate exists, say that no supersession candidate was
found.
3. **Apply the three-condition test** (from `ADR-FORMAT.md`):
- hard to reverse
- surprising without context
- the result of a real trade-off
If ANY condition is missing, STOP. Tell the user which condition fails and
that no ADR will be written. Do not write a weak ADR — over-recording is the
main failure mode.
4. **Draft on screen.** Keep it short. Show:
- the full ADR (frontmatter + title + 1-3 sentences, plus optional sections
only if they add value)
- `Supersession analysis`, listing the ADR candidates considered by file path
and the decision: supersedes ADR-NNNN, no supersession, or uncertain
Use optional frontmatter only when useful:
- `supersedes: ADR-NNNN` only when the new ADR replaces an existing ADR
- `superseded_by: ADR-NNNN` only on the older ADR being superseded
- `tags: [...]` for short retrieval categories
- `affected_components: [...]` for concrete boundaries/components
If the supersession is uncertain, ask the user before writing. Do not mutate
an existing ADR from a guess.
5. **Write without waiting.** Incorporate any edits the user gives. The
three-condition test in step 3 still applies.
6. **Apply with the deterministic helper.** Do not hand-edit ADR files,
supersession metadata, the index, or the `CLAUDE.md` pointer. Write a temporary JSON file
with this shape:
```json
{
"title": "Short title",
"body": "1-3 concise sentences in English.",
"supersedes": "ADR-0001",
"tags": ["storage"],
"affected_components": ["database"]
}
```
Include `supersedes`, `tags`, and `affected_components` only when useful.
Never include `superseded_by`; the helper owns the old ADR mutation. Run:
```bash
node "$HOME/.claude/skills/adr/scripts/apply-adr.mjs" --root . --input /tmp/adr-draft.json
```
If that path is unavailable, locate `scripts/apply-adr.mjs` next to this
skill. If the helper reports an existing ADR or index integrity
failure, stop and report the blocker; do not repair it in the ADR write path.
7. **Report.** Use the helper JSON output as source of truth. Name the file
written, the assigned number, whether any ADR was superseded, and whether the
`docs/adr/README.md` index was created, updated, or skipped.From Tonours/etabli, MIT.
More from Tonours/etabli
- 0.570score
react-doctor-100
Tonours
Runs react-doctor on React codebases and iteratively fixes diagnosed issues until the project reaches a verified score of 100 without shortcuts.
SkillAFreshness: AClaude Code +9 - 0.567score
employer-docs-writing
Tonours
Writes or rewrites pages for the employer Mintlify documentation (~/work/docs) in the corpus voice.
SkillAFreshness: AClaude Code +9 - 0.565score
thermo-nuclear-code-quality-review
Tonours
Runs an extremely strict maintainability review for abstraction quality, giant files, and spaghetti-condition growth.
SkillAFreshness: AClaude Code +9 - 0.561score
spec
Tonours
Writes or revises employer specs, technical Epics and design docs from the bundled templates.
SkillAFreshness: AClaude Code +9 - 0.561score
coolify
Tonours
Manages Coolify Cloud via its official CLI (servers, apps, databases, deployments, logs, env, backups).
SkillAFreshness: AClaude Code +9 - 0.561score
write-direct
Tonours
Writes or rewrites internal docs and messages in Anthony's direct, informal voice (French by default).
SkillAFreshness: AClaude Code +9 - 0.561score
code-quality
Tonours
Runs the bounded convention and stack-quality pass on a diff.
SkillAFreshness: AClaude Code +9 - 0.560score
implement
Tonours
Implements an existing READY PLAN.md without replanning.
SkillAFreshness: AClaude Code +9 - 0.560score
pr-council-review
Tonours
Runs a three-pass review of one PR (code-review + thermo-nuclear, then forest:validator as the council), dry-run by default.
SkillAFreshness: AClaude Code +9 - 0.560score
adversary
Tonours
Red-teams a PLAN.md, diff or design as a hostile challenger.
SkillAFreshness: AClaude Code +9 - 0.557score
plan-implement
Tonours
Plans, then implements the same task in one run once PLAN.md is READY.
SkillAFreshness: AClaude Code +9 - 0.556score
employer-backend-suite
Tonours
Routes employer server, agent-nodejs, BFF, MCP, OAuth/JWT, permissions v4, workflow execution and Zendesk work to existing KB notes.
SkillAFreshness: AClaude Code +9
Signals
- Stars
- 0
- Forks
- 0
- Open issues
- 0
- Last commit (this folder)
- Oct 23, 2026 (6 days ago)
From GitHub (fetched Oct 24, 2026).
Details
- License
- MIT
- Repository
- github.com/Tonours/etabli
- First seen
- Oct 24, 2026
- Last checked
- Oct 24, 2026
- Source
- GitHubgithub-graphql
More from Tonours
- 0.570score
react-doctor-100
Tonours
Runs react-doctor on React codebases and iteratively fixes diagnosed issues until the project reaches a verified score of 100 without shortcuts.
SkillAFreshness: AClaude Code +9 - 0.567score
employer-docs-writing
Tonours
Writes or rewrites pages for the employer Mintlify documentation (~/work/docs) in the corpus voice.
SkillAFreshness: AClaude Code +9 - 0.565score
thermo-nuclear-code-quality-review
Tonours
Runs an extremely strict maintainability review for abstraction quality, giant files, and spaghetti-condition growth.
SkillAFreshness: AClaude Code +9 - 0.561score
spec
Tonours
Writes or revises employer specs, technical Epics and design docs from the bundled templates.
SkillAFreshness: AClaude Code +9 - 0.561score
coolify
Tonours
Manages Coolify Cloud via its official CLI (servers, apps, databases, deployments, logs, env, backups).
SkillAFreshness: AClaude Code +9 - 0.561score
code-quality
Tonours
Runs the bounded convention and stack-quality pass on a diff.
SkillAFreshness: AClaude Code +9
Cite this page
adr skill. Agent Mart. https://agentmart.dev/skills/tonours-adr (last updated 2026-10-24)