fix(buzz-acp): strengthen agent communication rules in base prompt#1293
Merged
Conversation
Agents sometimes finish their turn without calling `buzz messages send`, leaving users with silence. Add a closing obligation rule to the General communication section so every turn that processes a user message must end with a visible reply. Co-authored-by: Will Pfleger <wpfleger@block.xyz> Signed-off-by: Will Pfleger <wpfleger@block.xyz>
…essages Agents were posting replies and task delegations to unrelated channels (e.g. #general) instead of the channel where the user tagged them. The [Context] block already supplies the correct channel UUID every turn but there was no rule requiring agents to use it for outbound messages. Add one sentence to the Threading section making this explicit: all replies and delegations go to the same channel where the agent was tagged, unless the user explicitly requests otherwise. Co-authored-by: Will Pfleger <pfleger.will@gmail.com> Signed-off-by: Will Pfleger <pfleger.will@gmail.com>
wpfleger96
changed the title
wpfleger96
added a commit
that referenced
this pull request
Jun 25, 2026
…1293) Signed-off-by: Will Pfleger <wpfleger@block.xyz> Signed-off-by: Will Pfleger <pfleger.will@gmail.com> Co-authored-by: npub1fgdl5qqnh3k3f2xkqrvt7cujalhm623x4s7fdjdj5yrtp5fzjl9qrjpucw <4a1bfa0013bc6d14a8d600d8bf6392efefbd2a26ac3c96c9b2a106b0d12297ca@sprout-oss.stage.blox.sqprod.co>
tellaho
pushed a commit
that referenced
this pull request
Jun 25, 2026
…ebar * origin/main: fix(sidebar): non-selectable channel names + copy/leave context menu actions (#1260) fix(runtime): sweep node wrapper processes hosting managed agent shims (#1296) fix(buzz-agent): follow symlinks when discovering skill directories (#1295) chore: add grab-emoji.sh to register Slack emoji in Buzz (#1292) Fix cross-pod membership notification fanout (#1291) fix(buzz-acp): strengthen agent communication rules in base prompt (#1293) chore(release): release Buzz Desktop version 0.3.34 (#1289) feat(desktop): refresh Agents tab live on inbound relay sync (#1256) fix(buzz-acp): inject Codex network allowlist for relay hostname at spawn time (#1287) Co-authored-by: Taylor Ho <taylorkmho@gmail.com> Signed-off-by: Taylor Ho <taylorkmho@gmail.com> # Conflicts: # desktop/src-tauri/src/commands/agent_models.rs
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
Two additions to
crates/buzz-acp/src/base_prompt.md:Send obligation — every turn that processes a user message MUST end with
buzz messages send. Prevents the silent failure mode where agents complete work via tool calls but never send a visible reply.Stay in the tagged channel — all replies and task delegations go to the same channel where the agent was tagged (channel UUID from
[Context]). Prevents agents from redirecting responses or assignments to unrelated channels (e.g.#general) when the user tagged them elsewhere.Both rules close gaps where agents had the right information in context but no explicit instruction on how to use it.