fix(wasm): run leak scan on pre-injection headers in channel callbacks

[nick-stebbings]

Summary

Run the leak detector on original WASM-provided headers before credential injection in channel HTTP callbacks. This is the companion fix to #791 (merged) which fixed the same bug in the tools wrapper.

Problem

The WASM channel host's http_request handler calls inject_credentials() to replace placeholder values (e.g. {SLACK_BOT_TOKEN}) with real secrets, then runs the leak scan. Host-injected credentials trigger the leak detector, blocking legitimate channel callbacks like Slack chat.postMessage.

Fix

Scan the raw WASM-provided headers (pre-injection) instead of the post-injection headers. WASM never sees the real token values, so the pre-injection state is the correct input for leak detection. Matches the existing pattern in src/tools/wasm/wrapper.rs (PR #791).

Test plan

• cargo test --lib passes (3150 tests)
• cargo clippy --all --all-features clean
• cargo fmt --check clean
• Manual: Slack bot responds without leak scan false positive on xoxb- token

🤖 Generated with Claude Code

[gemini-code-assist]

Summary of Changes

Hello, I'm Gemini Code Assist¹! I'm currently reviewing this pull request and will post my feedback shortly. In the meantime, here's a summary to help you and other reviewers quickly get up to speed!

This pull request addresses an issue where the leak detector in WASM channel HTTP callbacks would incorrectly flag host-injected credentials as potential leaks. By adjusting the timing of the leak scan to occur on the raw, pre-injection headers, the change ensures that only WASM-provided values are checked, thereby eliminating false positives and allowing legitimate channel callbacks to function correctly.

Highlights

• Prevented Leak Detector False Positives: The leak detector in WASM channel HTTP callbacks now scans the original headers provided by WASM before any host-side credential injection occurs. This resolves false positives caused by host-injected tokens (e.g., Slack bot tokens) triggering the detector.

Using Gemini Code Assist

The full guide for Gemini Code Assist can be found on our documentation page, here are some quick tips.

Invoking Gemini

You can request assistance from Gemini at any point by creating a comment using either /gemini <command> or @gemini-code-assist <command>. Below is a summary of the supported commands on the current page.

Feature	Command	Description
Code Review	/gemini review	Performs a code review for the current pull request in its current state.
Pull Request Summary	/gemini summary	Provides a summary of the current pull request in its current state.
Comment	@gemini-code-assist	Responds in comments when explicitly tagged, both in pull request comments and review comments.
Help	/gemini help	Displays a list of available commands.

Customization

To customize Gemini Code Assist for GitHub experience, repository maintainers can create a configuration file and/or provide a custom code review style guide (such as PEP-8 for Python) by creating and adding files to a .gemini/ folder in the base of the repository. Detailed instructions can be found here.

Limitations & Feedback

Gemini Code Assist may make mistakes. Please leave feedback on any instances where its feedback is incorrect or counter productive. You can react with 👍 and 👎 on @gemini-code-assist comments. If you're interested in giving your feedback about your experience with Gemini Code Assist for GitHub and other Google products, sign up here.

Footnotes

1. Review the Privacy Notices, Generative AI Prohibited Use Policy, Terms of Service, and learn how to configure Gemini Code Assist in GitHub here. Gemini can make mistakes, so double check it and use code with caution. ↩

[gemini-code-assist]

Code Review

This pull request addresses a potential security vulnerability in the WASM channel wrapper by ensuring that the leak detector scans the original WASM-provided headers before any credential injection occurs. This prevents false positives caused by host-injected credentials triggering the leak detector. The changes involve modifying the http_request function to extract and scan the raw headers before credential injection, aligning with the approach used in the tools wrapper. The test plan includes unit tests and manual testing to verify the fix.

[nick-stebbings]

Addressed Gemini review feedback in ec4267a:

• Malformed headers JSON: Replaced unwrap_or_default() with unwrap_or_else that logs a warning before falling back to empty headers. Malformed JSON from WASM is now visible in logs.
• Scope block: Kept intentionally — scopes the raw headers lifetime before credential injection to make the scan-before-inject ordering clear.

All checks pass (clippy, fmt, tests).

[ilblackdragon]

Review: fix(wasm): run leak scan on pre-injection headers in channel callbacks

Must-fix

1. Eliminate double-parse of headers_json — the code parses headers_json twice: once at line 340-341 for the existing flow, and again in the new block for the leak scan. The tools wrapper (src/tools/wasm/wrapper.rs:296-328) avoids this by scanning raw_headers BEFORE consuming it with into_iter(). Restructure to match:

let raw_headers = serde_json::from_str(&headers_json).unwrap_or_default();
// Leak scan on raw WASM-provided values
leak_detector.scan_http_request(&url, &header_vec, body.as_deref())?;
// THEN inject credentials
let headers = raw_headers.into_iter()
    .map(|(k, v)| (k.clone(), self.inject_credentials(&v, ...)))
    .collect();

2. Misleading comment — says "ORIGINAL WASM-provided values (before ANY credential injection)" but the URL is post-template-injection (injected_url). Fix to say "before host credential injection" or "pre-host-injection."

Should-fix

3. Migrate import — per CLAUDE.md, "new code should import from ironclaw_safety directly." Both the production code and test use crate::safety::LeakDetector. Migrate to ironclaw_safety::LeakDetector.

4. Remove unnecessary block scope — the { ... } around the leak scan provides no scoping benefit.

5. Remove raw_url_for_scan alias — it's just &url with no transformation.

[nick-stebbings]

Addressed all review feedback from @ilblackdragon in c594003, rebased onto latest staging (78e448df):

Must-fix (both resolved)

1. Double-parse eliminated — headers_json is now parsed once into raw_headers. Leak scan runs on the raw values via .iter(), then raw_headers.into_iter() feeds credential injection. Matches the tools wrapper pattern.

2. Comment fixed — Now reads "URL has template substitution applied (injected_url) but not yet host credential injection" (was incorrectly saying "before ANY credential injection").

Should-fix (all resolved)

3. Import migrated — Both production (line 54) and test (line 4798) imports changed from crate::safety::LeakDetector to ironclaw_safety::LeakDetector.

4. Block scope removed — Flat code, no unnecessary { }.

5. raw_url_for_scan alias removed — Uses &url directly.

All checks pass (cargo check, clippy, fmt, tests).

[zmanian]

Review -- APPROVE

Security fix is correct. Leak scan now runs on pre-injection headers, preventing false positives from host-injected credentials (e.g., Slack xoxb- tokens).

All of ilblackdragon's feedback addressed:

1. Double-parse eliminated -- single parse into raw_headers, consumed by into_iter() after scan
2. Comment updated to accurately describe pre/post-injection ordering
3. Import migrated to ironclaw_safety::LeakDetector
4. Block scope and URL alias removed
5. unwrap_or_default replaced with unwrap_or_else + tracing::warn

Single fix point in http_request host function correctly covers all WASM HTTP egress (other callbacks invoke guest code which calls back into http_request).

Minor suggestion (non-blocking)

Consider whether malformed headers_json should be a hard error rather than falling back to empty headers -- a WASM module producing unparseable JSON may be worth failing fast on.

[serrrfirat]

Approved. I resolved the staging merge conflict and re-checked the final diff against origin/staging; it stays scoped to the WASM channel wrapper leak-scan ordering. The rerun checks on head 7db495c3 are green, including formatting, clippy default/libsql/all-features, cargo-deny, no-panics, and regression enforcement.