fix(orchestrator): pass enriched task description to all subagents

Verification, execution optimizer, and final review agents were receiving
the raw taskDescription instead of effectiveTaskDescription (which includes
research context). This caused them to work with outdated/wrong prompts
when regenerating plans.

Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
This commit is contained in:
arkonandClaude Opus 4.5 committed 2026-01-28 06:19:03 +01:00
1 parent c17dc57fd9
commit 9eccd061b9
3 files changed
+6 -6

No files matched your search

+4 -4
View File
@@ -1088,10 +1088,10 @@ Check \`${caseDir}/ralph-wizard/research/result.json\` for:
onProgress?.('synthesis', 'Synthesizing subagent outputs...');
const synthesisResult = this.synthesizeResults(subagentResults);
// Phase 3: Verification
// Phase 3: Verification (use enriched task description from research)
onProgress?.('verification', 'Running verification subagent...');
const verificationResult = await this.runVerification(
taskDescription,
effectiveTaskDescription,
synthesisResult.items,
onProgress,
onSubagent
@@ -1112,7 +1112,7 @@ Check \`${caseDir}/ralph-wizard/research/result.json\` for:
try {
onProgress?.('execution-optimization', 'Running execution optimizer for Claude Code...');
executionResult = await this.runExecutionOptimizer(
taskDescription,
effectiveTaskDescription,
planWithReviews,
onProgress,
onSubagent
@@ -1146,7 +1146,7 @@ Check \`${caseDir}/ralph-wizard/research/result.json\` for:
try {
onProgress?.('final-review', 'Running final review for holistic validation...');
finalReviewResult = await this.runFinalReview(
taskDescription,
effectiveTaskDescription,
optimizedPlan,
onProgress,
onSubagent