Stats
Actions
Tags
From quantum-loop
Scans merged diffs at wave boundaries for divergent constants across stories, e.g., 'google' vs 'google-api-key' for the same concept. Logs findings, routes high-severity (3+ variants) to fix-stories.
How this agent operates — its isolation, permissions, and tool access model
Agent reference
quantum-loop:agents/orchestrator-modules/constant-scanThe summary Claude sees when deciding whether to delegate to this agent
**Activation:** runs at every wave boundary (no module-availability guard). Before any other wave-boundary check, scan the merged diff for divergent constants across stories — the Math-Research class where story A uses `'google'` and story B uses `'google-api-key'` for the same concept. Per-story review is blind to this because each story is correct in isolation. ```bash FINDINGS=$(bash lib/wav...
Activation: runs at every wave boundary (no module-availability guard).
Before any other wave-boundary check, scan the merged diff for divergent constants across stories — the Math-Research class where story A uses 'google' and story B uses 'google-api-key' for the same concept. Per-story review is blind to this because each story is correct in isolation.
# BASE is the wave's pre-merge SHA; HEAD is the post-merge tip
FINDINGS=$(bash lib/wave-boundary.sh scan "$WAVE_BASE_SHA" HEAD)
if [[ $(printf '%s' "$FINDINGS" | jq 'length') -gt 0 ]]; then
printf '[WAVE-BOUNDARY] Divergent constants detected:\n%s\n' "$FINDINGS" >&2
# Severity "high" (3+ variants) routes to a targeted fix-story;
# "medium" (2 variants) is logged + passed to ql-deep-review as input.
HIGH_COUNT=$(printf '%s' "$FINDINGS" | jq '[.[] | select(.severity=="high")] | length')
if [[ "$HIGH_COUNT" -gt 0 ]]; then
# Emit a fix-story via the same path as 3C.3 integration failures
echo "[WAVE-BOUNDARY] HIGH severity — routing to fix-story." >&2
fi
fi
Findings propagate into the deep-review context (Step 4B) even if non-blocking here.
npx claudepluginhub andyzengmath/quantum-loop --plugin quantum-loopScans git wave changes for semantic code clones (alpha-renamed duplicates) across JS/TS/Python/Go/Rust files via coarse file-level fingerprints. Advisory pre-screen logs groups for deep review.
Reviews an entire feature implementation for cross-task issues, identifying inconsistencies, duplication, dead code, integration gaps, and test coverage problems.
Deep code reviewer for logic bugs, data flow issues, edge cases, and pattern inconsistencies. Analyzes git diffs, traces flows across files, returns structured p0-p2 findings.