WCAG 2.1 AA audit — perceivable, operable, understandable, robust criteria. Deep-dive for /launch-check accessibility.
The English summary is being prepared. Showing the original description.
Make a technical decision with structured reasoning and create an Agent Decision Record (AgDR).
The English summary is being prepared. Showing the original description.
Run score Not tested
See how to installReview the instructions the agent will receive before you install.
Writing: read docs/skills/shared/writing-rule.md before you write a durable artifact.
Forces structured decision-making and creates an auditable Agent Decision Record (AgDR).
When /decide runs, activate the Tech Lead role — they own technical decisions within their domain. For decisions that cross the architecture-review threshold (new service, new tech stack, new external integration, major data model change), escalate to the Head of Engineering before creating the AgDR.
If the decision touches auth / crypto / secrets / PII, also involve the Security Auditor for sign-off on the security implications before finalising the choice.
See .claude/rules/role-triggers.md for the full activation protocol.
Extract the decision topic from $ARGUMENTS. If unclear, ask:
What technical decision do you need to make?
Identify decision-relevant context only:
Present 2–4 options in a table:
| Option | Pros | Cons |
|--------|------|------|
| Option A | … | … |
| Option B | … | … |
State the chosen option with justification.
Create file at {project-root}/docs/agdr/AgDR-{NNNN}-{slug}.md.
After writing the file, source .claude/hooks/_lib-review-markers.sh from the ops fork and run review_validate_body agdr <path-to-new-AgDR>. Repair any missing title or required section before reporting the decision. This local check does not approve the decision or write a review marker.
Important: AgDRs live in the current project's repository, not centralised. Each project has its own docs/agdr/ folder and its own ID sequence.
Resolve the AgDR template via the portfolio helper so adopter overrides win when present:
source "$(git rev-parse --show-toplevel)/.claude/hooks/_lib-read-config.sh"
source "$(git rev-parse --show-toplevel)/.claude/hooks/_lib-portfolio-paths.sh"
template=$(portfolio_resolve_template agdr.md) # → custom-templates/agdr.md if present, else templates/agdr.md
Single-fork adopters (no portfolio block) and adopters with no override fall straight through to templates/agdr.md. Adopters who want a customised AgDR shape drop their version at <private_repo>/custom-templates/agdr.md. See templates/README.md for the path-mirroring convention.
The skeleton:
---
id: AgDR-{NNNN}
timestamp: {ISO-8601: YYYY-MM-DDTHH:MM:SSZ}
agent: {current-agent-name or "claude"}
model: {model-id from environment}
trigger: {user-prompt | hook | automation}
status: executed
---
# {short title}
> In the context of {context}, facing {concern}, I decided {decision} to achieve {goal}, accepting {tradeoff}.
## Context
{Decision-relevant context only — 2–4 bullets}
## Options Considered
| Option | Pros | Cons |
|--------|------|------|
| … | … | … |
## Decision
Chosen: **{option}**, because {justification}.
## Consequences
- {consequence 1}
- {consequence 2}
## Artifacts
- {commit / PR links when available}
Use a filesystem lock while scanning and reserving the next ID. This prevents
two concurrent /decide runs from selecting the same number.
# Use the main worktree's shared Git directory so linked worktrees serialize
# allocation together.
git_common_dir=$(git rev-parse --path-format=absolute --git-common-dir)
ops_root=$(dirname "$git_common_dir")
lock_dir="${APEXYARD_AGDR_LOCK_DIR:-$ops_root/.claude/session}/agdr-id.lock"
while ! mkdir "$lock_dir" 2>/dev/null; do sleep 1; done
trap 'rmdir "$lock_dir" 2>/dev/null || true' EXIT
reservation_dir="${APEXYARD_AGDR_RESERVATION_DIR:-$ops_root/.claude/session/agdr-reservations}"
mkdir -p "$reservation_dir"
last=$(find docs/agdr "$reservation_dir" -maxdepth 1 -type f \( \
-name 'AgDR-[0-9][0-9][0-9][0-9]-*.md' -o -name 'AgDR-[0-9][0-9][0-9][0-9]' \) -print \
| sed -E 's#^.*/AgDR-([0-9]{4})(-.*)?$#\1#' | sort -n | tail -1)
next=$(printf '%04d' $((10#${last:-0} + 1)))
# Reserve the ID before writing the record. The reservation is shared by all
# linked worktrees and remains after a worktree is removed.
while ! (set -C; : > "$reservation_dir/AgDR-${next}") 2>/dev/null; do
next=$(printf '%04d' $((10#$next + 1)))
done
# Keep the reservation until the AgDR file is committed.
Keep the lock until the AgDR file is created. If the candidate exists after the scan, increment and check again. Do not delete a reservation after the AgDR is written; it prevents another branch from reusing the identifier.
A decision is a high-stakes moment. After drafting the AgDR, surface a one-line nudge offering to stress-test the call before committing:
Recorded AgDR-{NNNN}. Want Naqid (The Contrarian) to challenge this first?
Run `/challenge {topic}` — he steelmans it, then attacks hidden assumptions,
failure modes, and cheaper alternatives. Advisory — it never blocks.
Rules for the offer:
/challenge; the operator decides.This mirrors the advisory stance in .claude/rules/role-triggers.md
§ "Optional advisory offer" — see .claude/skills/challenge/SKILL.md and AgDR-0078.
Decision: {chosen option}
AgDR-{NNNN} created at docs/agdr/AgDR-{NNNN}-{slug}.md
Proceeding with: {brief action}
because clause is mandatory/challenge (opt-in,
advisory, never auto-run); update the AgDR if the verdict shifts the callPart of ApexYard — multi-project SDLC framework for Claude Code · MIT.
No reviews yet. Tell others what you thought.
No login needed. Share how it worked for you, ask a question, or note where it did not run.
No comments yet. Be the first to write one.
WCAG 2.1 AA audit — perceivable, operable, understandable, robust criteria. Deep-dive for /launch-check accessibility.
The English summary is being prepared. Showing the original description.
WCAG 2.2 AA audit — perceivable, operable, understandable, robust criteria. Deep-dive for /launch-check accessibility.
The English summary is being prepared. Showing the original description.
Browse / search / show / stats AgDRs across the portfolio — recalls "have we decided this before?".
The English summary is being prepared. Showing the original description.
Analytics audit — SDK config, event naming, funnel completeness, dashboards. Deep-dive for /launch-check analytics.
The English summary is being prepared. Showing the original description.
Record per-PR architecture-review approval for design-artifact PRs (required by the architecture gate). ONLY on an explicit per-PR architect "approved".
The English summary is being prepared. Showing the original description.
Record per-PR design-review approval (UI merge gate). ONLY on an explicit per-PR designer "approved".
The English summary is being prepared. Showing the original description.