chore: promote staging to main (2026-03-11 03:47 UTC)

[ironclaw-ci]

Auto-promotion from staging CI

Batch range: 55b5a462a2d2056cebc8cb4dec1680bff925e01a..369741fc60bf4ec1a28445c23d99db4a7f9c04c3
Promotion branch: staging-promote/369741fc-22935740447
Base: staging-promote/55b5a462-22934480277
Triggered by: Staging CI batch at 2026-03-11 03:47 UTC

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 6 issues:

1. [MEDIUM:75] Tool trait method returns WASM-specific type, creating circular dependency

The webhook_capability() method in the Tool trait returns Option<crate::tools::wasm::WebhookCapability>, forcing core tool infrastructure to depend on WASM-specific types. This violates CLAUDE.md: "Prefer generic/extensible architectures over hardcoding specific integrations."

https://github.com/anthropics/ironclaw/blob/486e417/src/tools/tool.rs#L335-L337

2. [MEDIUM:65] Header names hardcoded as strings with inconsistent defaults

Webhook signature header names are Option<String> fields with defaults scattered across src/webhooks/mod.rs. Multiple code paths produce different defaults.

https://github.com/anthropics/ironclaw/blob/486e417/src/tools/wasm/capabilities.rs#L258-L273

3. [LOW:60] Empty webhook capability configs pass parse-time validation

Schema allows all webhook fields to be None, validating only at runtime. Should fail at parse time (fail-fast principle).

https://github.com/anthropics/ironclaw/blob/486e417/src/tools/wasm/capabilities_schema.rs#L335-L358

4. [LOW:55] Late-binding shared routine engine slot creates potential race condition

shared_routine_engine_slot is created empty and lazily populated by the agent task. If a webhook request arrives before population, it returns 503.

https://github.com/anthropics/ironclaw/blob/486e417/src/main.rs#L281-L284

5. [LOW:50] No startup validation for secrets store availability

If webhook routes are registered but secrets store wasn't injected, requests fail at runtime. Should be caught at server startup.

https://github.com/anthropics/ironclaw/blob/486e417/src/webhooks/mod.rs#L712-L714

6. [LOW:50] Incomplete error context in webhook error handling

Tool webhook errors lack context about what the tool actually returned.

https://github.com/anthropics/ironclaw/blob/486e417/src/webhooks/mod.rs#L640-L650

Green lights: ✓ Constant-time signature verification ✓ No .unwrap() in production code ✓ Comprehensive test coverage ✓ Good separation of concerns ✓ Proper async/Arc usage

[claude]

Bug Scan Update:

Found one issue worth investigating:

[LOW:60] Test assertion weakened without explanation

In tests/e2e_advanced_traces.rs, the max tool iterations test assertion was changed from <= 4 to <= 8 while the test comment still states max_tool_iterations=3. Doubling the allowed tool calls could indicate:

• Legitimate behavioral change (agent now makes more calls)
• Test being weakened to mask a regression
• Copy-paste error

https://github.com/anthropics/ironclaw/blob/486e417/tests/e2e_advanced_traces.rs#L1202-L1206

This should be clarified in the PR description or git history.

[claude]

Performance & Production Review Update:

Found 2 MEDIUM severity performance/security issues:

[MEDIUM:70] Unbounded query parameters allow potential DoS attacks

Query parameters are collected into HashMap with no size limits. A malicious client could send thousands of query parameters to cause memory exhaustion.

https://github.com/anthropics/ironclaw/blob/486e417/src/webhooks/mod.rs#L99-L122

Recommend: Add query parameter count/size validation in webhook handlers.

[MEDIUM:68] Inefficient header HashMap allocation in hot path

On every webhook request, all HTTP headers are converted to HashMap by cloning both keys and values to Strings (2 allocations per header). For typical webhooks with 25-50 headers, this could be 50-100+ allocations per request.

https://github.com/anthropics/ironclaw/blob/486e417/src/webhooks/mod.rs#L157-L164

Recommend: Use references or BTreeMap, or defer HashMap creation until actually needed by the tool.

Additional LOW-severity findings:

• Unnecessary Arc clone of routine engine on every webhook
• String clones in error messages (should use &'static str)
• Unnecessary UTF-8 lossy conversion for request body

Overall: Good async/await patterns, proper timeouts and body limits. Production-ready with optimizations noted above.

[claude]

Security & Safety Review Update:

Found 2 MEDIUM severity security issues that should be addressed before merge:

[MEDIUM:85] Information disclosure via secret names in error messages

Error messages reveal which secrets are configured (e.g., 'Missing webhook secret github_webhook_secret'). Attackers can learn the naming convention and integrated services.

https://github.com/anthropics/ironclaw/blob/486e417/src/webhooks/mod.rs#L281-L323

Recommend: Return generic 'Authentication failed' messages. Log secret names internally only.

[MEDIUM:75] Potential timing side-channel in fallback header lookup

Header fallback lookup has variable timing before constant-time comparison. Attackers could measure response times to determine which header names are configured.

https://github.com/anthropics/ironclaw/blob/486e417/src/webhooks/mod.rs#L284-L292

Recommend: Always attempt both header lookups with constant timing.

Additional LOW-severity findings:

• Timestamp validation accepts zero if system clock before UNIX_EPOCH (bypass replay protection)
• No validation that retrieved secrets don't exceed maximum length (DoS risk)
• Inconsistent HTTP status codes (401 vs 404) could leak which tools exist

Positive: Good use of constant-time comparison (ct_eq) for secret validation, proper signature verification pattern.