feat(slack): implement on_broadcast and fix message tool hints

[serrrfirat]

Summary

• Implement on_broadcast in the Slack WASM channel (channels-src/slack/src/lib.rs): the agent can now send proactive messages to arbitrary Slack channels/users via the message tool. The user_id parameter carries the broadcast target (e.g. #general, C0123ABC, U0123); a leading # is stripped since the Slack API expects bare names or IDs.
• Fix message tool channel parameter hint (src/tools/builtin/message.rs): replace 'slack-relay' with 'slack' and clarify that "target" refers to Slack channels (not "Slack channel ID").
• Add 5 unit tests for the extracted helpers (resolve_broadcast_target, build_broadcast_payload).

Test plan

• cargo test in channels-src/slack/ passes (16 tests including 5 new)
• cargo clippy in channels-src/slack/ passes with zero warnings
• cargo fmt in channels-src/slack/ passes
• cargo clippy --all --benches --tests --examples --all-features passes in main repo
• cargo test passes in main repo (pre-existing e2e_spot_checks SIGABRT confirmed on staging)
• Manual: send a message via message tool targeting a Slack channel to verify end-to-end

🤖 Generated with Claude Code

[gemini-code-assist]

Code Review

This pull request implements the on_broadcast functionality for the Slack channel, enabling the agent to send proactive messages. It introduces helper functions for target resolution and payload construction, adds unit tests for these helpers, and updates the MessageTool schema description. A review comment suggests tracking these broadcasted messages in the active threads registry to ensure the agent can correctly handle subsequent replies in those threads.

[gemini-code-assist]

Proactive messages sent via on_broadcast should be tracked in the active threads registry, similar to how on_respond handles them. This ensures that if a user replies to the proactive message, the agent will be able to receive and process the reply even if it isn't explicitly mentioned (as the bot is now a participant in that thread).

Suggested change

	if let Some(ts) = response.thread_id.as_deref().or(slack_response.ts.as_deref()) {
	track_active_thread(&target, ts)?;
	}

[claude]

Code review

Found 6 issues:

1. [HIGH:75] Slack ID validation incomplete — only validates first two characters. A malicious target like C0\n<injected> would pass the looks_like_slack_id() check. While the immediate usage (JSON serialization) is safe, the weak validation creates a latent vulnerability if the target is ever used in string interpolation, SQL queries, or command construction elsewhere.

https://github.com/anthropics/ironclaw/blob/86c1590350c3e444fd20be1f17890447e57ac72f/channels-src/slack/src/lib.rs#L853-L860

Recommendation: Validate entire string with full alphanumeric check or use regex like ^[CUDGW][a-zA-Z0-9]+$.

2. [HIGH:75] Behavioral change in error propagation undocumented — on_respond() previously returned an error if track_active_thread() failed (fatal). Now errors are logged as warnings and the function succeeds. This is a semantic change that affects thread tracking reliability and should be documented with an explanation of why thread-tracking failures are now non-fatal.

https://github.com/anthropics/ironclaw/blob/86c1590350c3e444fd20be1f17890447e57ac72f/channels-src/slack/src/lib.rs#L276-L283

3. [MEDIUM:75] Inconsistent error handling across channels — Slack now silently logs thread-tracking failures while other channels (e.g., Telegram) propagate errors as fatal. This creates inconsistent semantics in the multi-channel abstraction. Either document that thread tracking is optional for Slack, or align the behavior.

https://github.com/anthropics/ironclaw/blob/86c1590350c3e444fd20be1f17890447e57ac72f/channels-src/slack/src/lib.rs#L276-L283

4. [MEDIUM:50] Redundant prune_active_threads() call — track_active_thread() calls prune before and after inserting a new thread, but no mutations happen between the two calls. The second prune is redundant and causes O(256) duplicate work per message.

https://github.com/anthropics/ironclaw/blob/86c1590350c3e444fd20be1f17890447e57ac72f/channels-src/slack/src/lib.rs#L609-L616

Recommendation: Remove the second prune_active_threads() call after the insert.

5. [LOW:75] Logging level misalignment with CLAUDE.md — Per CLAUDE.md, warn! output appears in the REPL and corrupts the terminal UI. Internal diagnostics like thread-tracking failures should use LogLevel::Debug, not Warn.

https://github.com/anthropics/ironclaw/blob/86c1590350c3e444fd20be1f17890447e57ac72f/channels-src/slack/src/lib.rs#L278-L281

6. [LOW:50] Missing design documentation — on_broadcast() rejects non-ID targets (e.g., #general) with an error, but there's no comment explaining why names are prohibited (e.g., immutability, security, avoiding name-resolution latency). Add a brief comment documenting the design decision.

https://github.com/anthropics/ironclaw/blob/86c1590350c3e444fd20be1f17890447e57ac72f/channels-src/slack/src/lib.rs#L309-L315