This is a minimal, low-risk feature to wire the unused RUFLO_COST_BUDGET_MULTIPLIER config into three hive execution functions. The multiplier allows cost-aware agent scaling (e.g., 2.0 = double agents up to hard cap; 0.5 = half agents).

• Scope: 1 file, 3 locations, ~7 lines per location
• Complexity: Low — simple bash math with clamping
• Risk: Minimal — fully backward-compatible (unset = no-op)
• Dependencies: None — uses already-exported env var

What is the minimum viable change?

• Add 7 lines of bash logic after each max_agents assignment in the 3 hive functions
• Logic: if multiplier is set, apply it with clamping to [1, base_max]

Implicit requirements:

• Multiplier must be optional (unset → current behavior preserved)
• Result must clamp to [1, hard_cap] (prevent 0 agents or exceeding limit)
• Handle floating-point multipliers (0.5, 2.0) safely using awk

Acceptance criteria:

• ✅ RUFLO_COST_BUDGET_MULTIPLIER=2.0 scales up to hard cap
• ✅ RUFLO_COST_BUDGET_MULTIPLIER=0.5 reduces by 50% (min 1)
• ✅ Unset multiplier = current behavior (backward compatible)
• ✅ scripts/sw-ruflo-adapter-test.sh passes
• ✅ npm test passes

Rationale for Alternative 1: Simplest approach, exactly matches issue specification, minimal code duplication is acceptable for ~7 lines of straightforward bash.

Depends on:

• RUFLO_COST_BUDGET_MULTIPLIER (already exported at line 138)
• awk (standard on all platforms)
• Bash parameter expansion (standard)

Depended on by:

• Callers of ruflo_execute_review(), ruflo_execute_quality(), ruflo_execute_audit()
• Pipeline stages: review, compound_quality, audit

Circular dependencies: None — pure config → behavior flow.

Can be simpler? No — logic is already minimal (one if + one awk). Reuse infrastructure? Yes — using already-exported env var. Simpler approach work? No — needs clamping for edge cases.

# After the max_agents/cq_agents assignment:
if [[ -n "${RUFLO_COST_BUDGET_MULTIPLIER:-}" ]]; then
 local _default_max="$max_agents"
 max_agents=$(awk -v d="$_default_max" -v m="${RUFLO_COST_BUDGET_MULTIPLIER}" \
 'BEGIN{v=int(d*m); print (v<1?1:(v>d?d:v))}' 2>/dev/null || echo "$max_agents")
fi

Semantic: clamped_agents = max(1, min(base_max, floor(base_max ×ばつ multiplier)))

• Insert multiplier logic after line 816 (local max_agents=...)
• Variable: max_agents

• Insert multiplier logic after line 961 (local cq_agents=...)
• Variable: cq_agents (note: different variable name)

• Insert multiplier logic after line 1069 (local max_agents=...)
• Variable: max_agents

• Set RUFLO_COST_BUDGET_MULTIPLIER=2.0, RUFLO_MAX_AGENTS=2
• Expect: agent count scales to hard cap (e.g., 4)

• Set RUFLO_COST_BUDGET_MULTIPLIER=0.5, RUFLO_MAX_AGENTS=10
• Expect: agent count = 5

• Set RUFLO_COST_BUDGET_MULTIPLIER=0.1, RUFLO_MAX_AGENTS=10
• Expect: agent count = 1 (clamped)

• Set RUFLO_COST_BUDGET_MULTIPLIER="invalid", RUFLO_MAX_AGENTS=4
• Expect: agent count = 4 (fallback)

• Unset RUFLO_COST_BUDGET_MULTIPLIER, RUFLO_MAX_AGENTS=4
• Expect: agent count = 4 (unchanged)

• Execute: scripts/sw-ruflo-adapter-test.sh
• Verify: all tests pass (existing + new)

• Execute: npm test
• Verify: no regressions in other modules

• Task 1: Add multiplier logic to ruflo_execute_review() after line 816
• Task 2: Add multiplier logic to ruflo_execute_quality() after line 961 (note: use cq_agents variable)
• Task 3: Add multiplier logic to ruflo_execute_audit() after line 1069
• Task 4: Add test case for multiplier=2.0 (scale up)
• Task 5: Add test case for multiplier=0.5 (scale down)
• Task 6: Add test case for multiplier=0.1 (clamp to min)
• Task 7: Add test case for multiplier=invalid (fallback)
• Task 8: Add test case for unset multiplier (backward compat)
• Task 9: Run unit test suite (scripts/sw-ruflo-adapter-test.sh)
• Task 10: Run full test suite (npm test)
• Task 11: Verify backward compatibility (no multiplier set)
• Task 12: Commit changes with issue reference

1. Execute scripts/sw-ruflo-adapter-test.sh
2. All 5 new tests + existing tests must pass
3. Verify no existing tests are broken

1. Execute npm test (full suite)
2. Verify no regressions in other modules

# Test 1: Scale up
export RUFLO_COST_BUDGET_MULTIPLIER=2.0 RUFLO_MAX_AGENTS=2
scripts/sw-ruflo-adapter-test.sh
# Test 2: Scale down
export RUFLO_COST_BUDGET_MULTIPLIER=0.5 RUFLO_MAX_AGENTS=10
scripts/sw-ruflo-adapter-test.sh
# Test 3: Backward compatibility
unset RUFLO_COST_BUDGET_MULTIPLIER
export RUFLO_MAX_AGENTS=4
scripts/sw-ruflo-adapter-test.sh

• All 3 functions modified with multiplier logic
• All 5 new test cases added and passing
• Unit test suite passes: scripts/sw-ruflo-adapter-test.sh
• Full test suite passes: npm test
• No regressions in existing functionality
• Multiplier=2.0 scales agents up to hard cap ✓
• Multiplier=0.5 reduces agents by 50% ✓
• Multiplier=0.1 clamps to min 1 agent ✓
• Invalid multiplier gracefully ignored ✓
• Unset multiplier preserves backward compatibility ✓
• Commit message references issue #414
• Branch ready for PR/review

Approach: Add 7-line multiplier logic to each of the 3 hive execution functions, with clamping to [1, hard_cap]. Fully backward-compatible (unset multiplier = no-op).

Files: 2 files, ~40 lines total (code + tests)

Risk: Minimal — env-var driven, no new dependencies, backward-compatible

Effort: ~30 minutes (code + tests + validation)