chore: promote staging to staging-promote/7234700c-23635804857 (2026-03-27 08:13 UTC)

[ironclaw-ci]

Auto-promotion from staging CI

Batch range: 7234700c78d985ddc872721bc2a7130eeaa0b8c3..2f4eb08613cefff1af8b7b1a475fda00c84dd855
Promotion branch: staging-promote/2f4eb086-23637233594
Base: staging-promote/7234700c-23635804857
Triggered by: Staging CI batch at 2026-03-27 08:13 UTC

Commits in this batch (2):

• 30db07c fix: require Feishu webhook authentication (fix: require Feishu webhook authentication #1638)
• 2f4eb08 fix: sanitize tool error results before llm injection (fix: sanitize tool error results before llm injection #1639)

Current commits in this promotion (2)

Current base: staging-promote/7234700c-23635804857
Current head: staging-promote/2f4eb086-23637233594
Current range: origin/staging-promote/7234700c-23635804857..origin/staging-promote/2f4eb086-23637233594

• 30db07c fix: require Feishu webhook authentication (fix: require Feishu webhook authentication #1638)
• 2f4eb08 fix: sanitize tool error results before llm injection (fix: sanitize tool error results before llm injection #1639)

Auto-updated by staging promotion metadata workflow

Waiting for gates:

• Tests: pending
• E2E: pending
• Claude Code review: pending (will post comments on this PR)

---

Auto-created by staging-ci workflow

[claude]

Code review

Found 4 issues:

1. [CRITICAL:95] Type erasure in process_tool_result() signature — accepting Result<String, impl Display> loses type information and prevents future error recovery or pattern matching. Consider using a trait object or keeping specific error types (ToolError, String).

https://github.com/anthropics/ironclaw/blob/41468f39331f7a611b9a38f61b43384333d56b78/src/tools/execute.rs#L828-L843

let (result_content, tool_message) = preflight_rejection_tool_message(
    self.agent.safety(),
    &tc.name,
    &tc.id,
    &error_msg,
);

2. [HIGH:95] Workspace I/O on hot path — is_authenticated_webhook() reads verification token from workspace storage on every webhook request, adding synchronous I/O latency. For high-frequency webhooks, consider caching the token or pre-loading during initialization.

https://github.com/anthropics/ironclaw/blob/41468f39331f7a611b9a38f61b43384333d56b78/channels-src/feishu/src/lib.rs#L389-L397

let configured_token =
    channel_host::workspace_read(VERIFICATION_TOKEN_PATH).filter(|token| !token.is_empty());
if !is_authenticated_webhook(
    req.secret_validated,
    configured_token.as_deref(),
    request_verification_token(&event),
) {

3. [MEDIUM:85] LazyLock duplication across execution paths — process_tool_result() creates a SafetyLayer, and process_builder_tool_result() creates another. Both are static singletons with identical initialization. Consider centralizing in the SafetyLayer module to avoid duplication.

https://github.com/anthropics/ironclaw/blob/41468f39331f7a611b9a38f61b43384333d56b78/src/tools/builder/core.rs#L1-L50

static SAFETY: LazyLock<SafetyLayer> = ...

4. [MEDIUM:80] Inconsistent error message format across code paths — New code formats errors via process_tool_result() as "Tool 'X' failed: ...", but legacy rebuild_chat_messages_from_db() passes errors through unchanged. This creates a silent contract that new code must always use the centralized function.

https://github.com/anthropics/ironclaw/blob/41468f39331f7a611b9a38f61b43384333d56b78/src/agent/thread_ops.rs#L1907-L1920

let content = if let Some(err) = c.get("error").and_then(|v| v.as_str())
{
    // Both wrapped (new) and legacy (plain) errors pass
    // through as-is. Legacy errors are already descriptive
    // (e.g. "Tool 'http' failed: timeout"), so no prefix needed.
    err.to_string()