fix: prevent Discord NO_REPLY bot loops

[JeffreyDavidson]

Summary

• Drop Discord bot messages whose content is exactly NO_REPLY before they can enter the agent loop
• Exclude bot NO_REPLY sentinel messages from Discord channel history backfill
• Suppress literal NO_REPLY platform responses at delivery time
• Add regression coverage for the Discord bot filter behavior

Why

When bot-to-bot routing is enabled with DISCORD_ALLOW_BOTS=mentions, a bot NO_REPLY sentinel can be treated as a real prompt. If the receiving agent then intentionally stays silent, the gateway can surface empty-response retries and create noisy bot-to-bot loops. NO_REPLY should behave as a control/silence token, not a user prompt.

Verification

• ./venv/bin/python -m pytest tests/gateway/test_discord_bot_filter.py tests/gateway/test_discord_bot_auth_bypass.py -q

[teknium1]

Thanks for the focused three-layer investigation. Current main has since centralized the outbound part of this behavior: gateway/response_filters.py:56-79 recognizes successful exact NO_REPLY turns, gateway/run.py:12146-12156 suppresses their delivery after persistence, and gateway/stream_consumer.py:606-620 covers streaming output.

Problems

• The PR edits the removed gateway/platforms/discord.py; the active Discord implementation is now plugins/platforms/discord/adapter.py, and GitHub reports this branch as conflicting.
• The added test duplicates the proposed conditional instead of exercising the adapter, so it cannot verify the live callback after the platform migration.

Suggested changes

• If the remaining inbound/history hygiene is desired, port it narrowly to plugins/platforms/discord/adapter.py:1130 and :5050, with adapter-level tests.
• Please account for the documented constraint at website/docs/user-guide/messaging/discord.md:304-307: Hermes-to-Hermes Discord reply loops are an unsupported topology; the remaining change should be justified as trusted-external-bot hygiene rather than a circuit breaker.

Automated hermes-sweeper review.

[GottZ]

This was generated by AI during triage.

Summary

Two PRs address the Discord NO_REPLY loop through inbound filtering, history exclusion, and outbound suppression. Their Discord changes substantially overlap, but #29932 is focused while #29942 also bundles multiple unrelated fixes; moreover, current main already centralizes outbound NO_REPLY suppression, and both diffs target the removed pre-migration Discord module.

Related pull requests

• #29932 related — (+35/-0) — keep open for a narrow rework, not merge as-is: the diff directly filters bot-authored exact NO_REPLY messages, excludes them from history, and suppresses outbound delivery, but it edits the removed gateway/platforms/discord.py and its test reproduces the condition rather than exercising the active adapter. Consistent with the contributor keep_open review on #29932, port only the still-needed inbound/history hygiene to plugins/platforms/discord/adapter.py with adapter-level tests; drop the now-duplicated outbound change.
• #29942 [closed] duplicate — (+399/-16) — closed, relevant as an overlapping reference implementation but not a merge candidate: its Discord portion implements the same three layers as #29932, while most of the diff contains unrelated tool-result serialization, provider, compression, runtime-provider, and skill-sync changes. The contributor review explicitly required these concerns to be split, so its focused Discord logic may inform #29932 but the closed PR should not be reopened or merged.

Duplicates

#29932 and the Discord-specific portion of #29942 are substantially duplicate implementations of inbound NO_REPLY filtering, history exclusion, and outbound suppression; #29942 additionally contains extensive unrelated changes.

Suggested consolidation

Do not merge either branch as-is. Consolidate the remaining work in #29932 by retaining it open, rebasing onto current main, porting only justified inbound/history filtering to the active Discord adapter with live-path tests, and omitting outbound suppression already centralized on main; keep #29942 closed as the superseded, unfocused duplicate reference.

Complex graph

flowchart LR
    classDef open fill:#dbeafe,stroke:#1d4ed8,color:#1e3a8a
    classDef merged fill:#dcfce7,stroke:#15803d,color:#14532d
    classDef closed fill:#e5e7eb,stroke:#6b7280,color:#1f2937
    classDef unverified fill:#f3f4f6,stroke:#9ca3af,color:#374151
    classDef best stroke-width:3px,stroke:#b45309
    classDef target stroke-width:3px,stroke:#4338ca
    subgraph Dup29932 ["PRs duplicating each other"]
        P29932["PR #29932 (open)"]
        P29942["PR #29942 (closed)"]
    end
    class P29932 open
    class P29942 closed
    class P29932 target
    click P29932 "https://github.com/NousResearch/hermes-agent/pull/29932"
    click P29942 "https://github.com/NousResearch/hermes-agent/pull/29942"

Loading

Graph: solid arrow = fixes / best fix, dashed arrow = partial or unverified (see edge label); boxed group = PRs duplicating each other; amber border = best fix; indigo border = target; gray node = closed or no verify verdict yet (state tag in the node label).

Cross-PR triage: Reviewed 2 pull requests and 0 issues in this complex. Each diff was read against this issue; Assessment working set: 31 kB of PR diffs, 2 kB of issue/PR text, 2 kB of discussion (3 comments), 1 verify verdict. verdicts reflect diff content, not PR titles. Part of an automated triage batch.