fix(Slack): resolve Slack channels by raw ID and enumerate joined channels

[hhuang91]

What does this PR do?

When asking Hermes to send a Slack message to a channel other than the
current one, two failure modes appeared:

1. send_message(target='slack:<channel_id>') returned Could not resolve '<channel_id>' on slack, even though the bot was a member of the channel and the ID was valid.
2. As a fallback, the agent would retry with the bare target='slack'
   form and silently post to the platform's home channel instead of
   the requested one.

In-channel and DM conversations worked fine — the bug only surfaced
on outbound cross-channel sends.

Root cause

Three compounding gaps in the send_message → channel_directory
path, all specific to Slack:

1. tools/send_message_tool.py::_parse_target_ref had explicit
   handlers for Telegram/Discord/Feishu/Weixin/Matrix/phone IDs, but
   no Slack branch. Slack's uppercase-alphanumeric IDs (C…, G…,
   D…, U…, W…) failed the existing target_ref.lstrip("-").isdigit()
   check and fell through with is_explicit=False, forcing the value
   into the name-resolver.
2. gateway/channel_directory.py::resolve_channel_name only
   matched against channel["name"], never channel["id"]. So even
   when a channel was in the directory, supplying its raw ID didn't
   resolve.
3. gateway/channel_directory.py::_build_slack never actually
   called the Slack web API — it just delegated to
   _build_from_sessions, so the directory only knew about channels
   the bot had received inbound messages in. Channels the bot had
   joined but never been talked to in were absent from
   action='list' and unaddressable by name.

The "redirected to home channel" symptom was the model's own
fallback: after the resolver error, it retried with bare
target='slack', which the schema documents as "uses home channel"
— compounding the visible misbehavior.

Fix

• _parse_target_ref recognizes Slack channel/user IDs as explicit
  targets via a precompiled regex ([CGDUW][A-Z0-9]{8,}), so they
  bypass name resolution entirely.
• resolve_channel_name adds a case-sensitive raw-ID match step
  before the existing case-insensitive name match. Defense in depth
  for any platform: a raw ID always resolves to itself if present in
  the directory.
• _build_slack now calls users.conversations against each
  workspace's AsyncWebClient, paginates via
  response_metadata.next_cursor, and merges in DM entries from
  session history. Per-workspace errors are isolated. Requires
  channels:read and groups:read Slack scopes for full
  enumeration; missing scopes degrade gracefully for that workspace
  alone.
• build_channel_directory becomes async (Slack web calls require
  it). Two async-context callers in gateway/run.py are awaited;
  the cron-ticker thread call bridges via
  asyncio.run_coroutine_threadsafe(...).result(timeout=30).

Why this approach

• Surgical, not architectural. The Slack ID parsing branch
  matches the pattern every other platform already uses in
  _parse_target_ref — it's the smallest change that brings Slack to
  parity.
• ID-match-before-name in resolve_channel_name is platform-
  agnostic: it costs one cheap loop and prevents this entire class of
  bug from recurring on any future platform whose ID format isn't
  caught by _parse_target_ref.
• users.conversations over conversations.list because it
  scopes to channels the bot is actually a member of, requires
  fewer permissions, and matches how Slack itself recommends bots
  enumerate their own workspace presence.
• Async over sync-with-requests because the adapter already
  owns the typed AsyncWebClients with multi-workspace tokens, proxy
  config, and retry semantics — reusing them avoids a parallel
  HTTP path and keeps token plumbing in one place.

Tests

24 new test cases:

• tests/tools/test_send_message_tool.py::TestParseTargetRefSlack (7) —
  public/private/DM/user IDs, whitespace, lowercase/short rejection,
  isolation from other platforms.
• tests/gateway/test_channel_directory.py::TestResolveChannelName::test_id_match_takes_precedence_over_name (1) —
  raw IDs resolve to themselves, lowercase still falls through to
  name matching.
• tests/gateway/test_channel_directory.py::TestBuildSlack (7) —
  no-clients fallback, single-page list, pagination via cursor,
  per-workspace error isolation, session-DM merge with dedup,
  missing-id/name skip, ok=False handling.

Existing test_failed_write_preserves_previous_cache updated to
asyncio.run(build_channel_directory({})).

Full test run: 118 passed (1 pre-existing Windows /tmp/ failure
unrelated to this change).

Related Issue

Fixes #15927

Type of Change

• 🐛 Bug fix (non-breaking change that fixes an issue)
• ✨ New feature (non-breaking change that adds functionality)
• 🔒 Security fix
• 📝 Documentation update
• ✅ Tests (adding or improving test coverage)
• ♻️ Refactor (no behavior change)
• 🎯 New skill (bundled or hub)

Changes Made

Three fixes:

• _parse_target_ref recognizes Slack IDs (C/G/D/U/W prefix) as explicit targets so the name-resolver is bypassed entirely.
• resolve_channel_name tries a case-sensitive raw-ID match before the existing name match, so any platform's IDs resolve cleanly.
• _build_slack now actually calls users.conversations against each workspace's AsyncWebClient (paginated), instead of only returning session-history entries. This populates the directory with public and private channels the bot has joined, so action='list' shows them and they can also be addressed by name. Errors from one workspace don't block others.

How to Test

1. Add Hermes to at least two Slack channels (e.g. #general, and #social)
2. In channel 1 (#general) ask Hermes to send a message (e.g. "hello world") in channel 2 (#social)
3. Hermes should say hello world in channel 2, while reporting message successfully sent in channel 1.

Checklist

Code

• I've read the Contributing Guide
• My commit messages follow Conventional Commits (fix(scope):, feat(scope):, etc.)
• I searched for existing PRs to make sure this isn't a duplicate
  • There is a existing PR fix(tools): treat Slack channel IDs as explicit send_message targets #2897 , but that one has been closed without merging.
• My PR contains only changes related to this fix/feature (no unrelated commits)
• I've run pytest tests/ -q and all tests pass
• I've added tests for my changes (required for bug fixes, strongly encouraged for features)
• I've tested on my platform: Ubuntu 24.04

Documentation & Housekeeping

• I've updated relevant documentation (README, docs/, docstrings) — or N/A
• I've updated cli-config.yaml.example if I added/changed config keys — or N/A
• I've updated CONTRIBUTING.md or AGENTS.md if I changed architecture or workflows — or N/A
• I've considered cross-platform impact (Windows, macOS) per the compatibility guide — or N/A
• I've updated tool descriptions/schemas if I changed tool behavior — or N/A

[alt-glitch]

Fixes #15927. Related to #15939 which addresses the same Slack channel ID parsing issue — this PR is more comprehensive (also fixes channel directory and enumeration).

[teknium1]

Merged via #16198 — your commit was cherry-picked onto current main with your authorship preserved (d889ad7). Thanks for both filing the issue with a clear three-layer fix plan AND following through with the full implementation. The users.conversations enumeration is the real fix for this class of bug — name-based addressing won't work reliably without it.

Small tweak in a follow-up commit: tightened the regex from [CGDUW] to [CGD] and salvaged that from @briandevans's #15939, because chat.postMessage rejects user IDs directly (U/W) — they'd fail at the API level, which is worse than falling through to name resolution and returning a clear error.
#16198