fix(scheduler): log WriteFile error + EM state run 10 — closes #65

[jpleva91]

Summary

• Fixes os.WriteFile silently discarding errors in executeAgent() — closes bug: scheduler.go silently ignores os.WriteFile error when writing agent run log #65
• Adds fmt.Printf warning when log write fails so audit log failures are visible
• Includes EM state update for run 10

What changed

internal/scheduler/scheduler.go line 242 (one-line fix):

// Before (silent discard)
os.WriteFile(logPath, []byte(logContent), 0o644)

// After (logged error)
if err := os.WriteFile(logPath, []byte(logContent), 0o644); err != nil {
    fmt.Printf("[scheduler] ⚠ %s: failed to write log: %s\n", agent.Name, err)
}

Why this matters

ShellForge is a governance tool — silent audit log loss is a P2 reliability issue. If outputs/logs/ doesn't exist, disk is full, or permissions are wrong, run telemetry was silently discarded with no indication.

Test plan

• go build ./cmd/shellforge/ passes
• go test ./... — 25/25 pass (governance, intent, normalizer)
• No call-site changes

EM run 10 actions (this run)

• Closed stale PR chore(squad): EM state update — run 8 (2026-03-30) #91 (DIRTY merge conflict, superseded by PR chore(squad): EM state update — run 9 (2026-03-31) #94)
• Fixed bug: scheduler.go silently ignores os.WriteFile error when writing agent run log #65 in this PR
• PR budget: 3/3 — human review of fix(main): log errors from run() helper — closes #51 #93 + chore(squad): EM state update — run 9 (2026-03-31) #94 needed to unblock queue

🤖 Generated with Claude Code