1. Search for a canonical root-cause tracker; add a new occurrence there when applicable
2. Choose the correct template and report consequence plus release context, not root cause or implementation
3. Submit → issue gets status:needs-review automatically
4. A maintainer triages and assigns an investigator, who records each same-root occurrence on the canonical tracker before duplicate closure
5. Use proportional evidence: scope/checks for trivial changes; current-remote-main SHA, reproduction, violated invariant/root cause, and regression test plan for bugs; approved design/non-goals and independent review plan for high-risk changes
6. A maintainer replaces the current status with status:approved after accepting that evidence; any real assignee satisfies ownership, not necessarily the PR author

gh issue create --template "bug_report.yml" \
  --title "fix(store): duplicate observations on concurrent saves" \
  --body "
### Observed Consequence
When two agents save observations concurrently, duplicates are created.

### Steps to Reproduce
1. Start two engram instances pointing to the same DB
2. Both save an observation with the same title simultaneously
3. Query observations — duplicates appear

### Expected Behavior
The second save should upsert, not insert a duplicate.

### Actual Behavior
Two identical observations exist with different IDs.

### Operating System
macOS

### Release Version
0.3.1

### Agent / Client
Claude Code

### Relevant Logs
\`\`\`
UNIQUE constraint failed: observations.title
\`\`\`
"

gh issue create --template "feature_request.yml" \
  --title "feat(cli): add --json flag to mem search" \
  --body "
### Observed Consequence
When scripting with engram, parsing human-readable search results is fragile.

### Desired Outcome
Scripts can consume search results without parsing human-readable output.

### Release Context
not release-specific

### Affected Area
CLI (commands, flags)

### Workarounds and Constraints
Using \`jq\` to parse the current output is unreliable because the format is unstructured.
"

1. Search for a canonical tracker, triage the consequence, and assign an investigator
2. For bugs, reproduce against exact current remote main and record SHA; identify the first violated repository-owned invariant
3. Cluster only same-root/contract occurrences; keep distinct roots separate
4. Accept evidence proportional to risk using CONTRIBUTING.md; a short issue explanation suffices, no separate design document by default
5. If accepted → replace the current status with status:approved; otherwise explain and close if needed
6. Require approval and a real assignee for new implementation; existing PRs must comply before merge, not before backlog-wide activation

Is it a bug?                    → Use Bug Report template
Is it a new feature/improvement? → Use Feature Request template
Is it a general question?       → Use Discussions, NOT issues
Is it a tracked question?       → Use the Tracked Question template
Is it a possible duplicate?     → Replace the current status with status:possible-duplicate
Is it a confirmed duplicate?    → Close it, replace evaluation status with status:wontfix, then add resolution:duplicate

# Search existing issues before creating to keep canonical evidence together
gh issue list --search "keyword"

# Create bug report
gh issue create --template "bug_report.yml" --title "fix(scope): description"

# Create feature request
gh issue create --template "feature_request.yml" --title "feat(scope): description"

# Create documentation improvement
gh issue create --template "docs_improvement.yml" --title "docs(scope): description"

# Create tracked question
gh issue create --template "tracked_question.yml" --title "question(scope): description"

# Maintainer: approve an issue
gh issue edit <number> --remove-label "status:needs-review" --add-label "status:approved"

# Maintainer: add priority
gh issue edit <number> --add-label "priority:high"

# Maintainer: replace every active status while evaluating a possible duplicate
set -euo pipefail

other_statuses="$(gh issue view <number> --json labels --jq '.labels[].name | select(startswith("status:") and . != "status:possible-duplicate")')"
status_args=(--add-label "status:possible-duplicate")
if [[ -n "$other_statuses" ]]; then
  status_args+=(--remove-label "${other_statuses//$'\n'/,}")
fi
gh issue edit <number> "${status_args[@]}"

updated_statuses="$(gh issue view <number> --json labels --jq '.labels[].name | select(startswith("status:"))')"
if [[ "$updated_statuses" != "status:possible-duplicate" ]]; then
  echo "status replacement could not be confirmed; stop without retrying" >&2
  exit 1
fi

# Maintainer: close a confirmed duplicate, then replace evaluation status with terminal status and resolution
gh issue close <number> --reason "not planned" --comment "Closing as duplicate of #<canonical>."
gh issue edit <number> --remove-label "status:possible-duplicate" --add-label "status:wontfix,resolution:duplicate"