beads,roles: emit BeadClaimed event; add bead loop to worker roles

BeadClaimed fires on successful claim with source=beads/<mount> and
data={bead_id, assignee, mount}.

Worker roles (developer, devops, pkgmgr, researcher, tester, author)
now have a Bead Loop section: wait_for_bead.sh → claim → work →
complete → repeat. Reviewer role is spawned by workers, not looping.

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

EVENTS.md

@@ -34,6 +34,7 @@ Each line is a JSON object:
 - `BotRecv` - Message received by bot
 - `BotSend` - Message sent by bot
 - `BeadReady` - A bead transitioned to open/ready; `source` is `beads/<mount>`, `data` is full bead JSON including comments
+- `BeadClaimed` - A bead was claimed by an agent; `source` is `beads/<mount>`, `data` is `{"bead_id","assignee","mount"}`
 
 ## Consuming Events
 

internal/eventbus/bus.go

@@ -19,7 +19,8 @@ const (
 	EventUserSend    = "UserSend"
 	EventBotRecv     = "BotRecv"
 	EventBotSend     = "BotSend"
-	EventBeadReady   = "BeadReady" // a bead transitioned to open/ready
+	EventBeadReady   = "BeadReady"   // a bead transitioned to open/ready
+	EventBeadClaimed = "BeadClaimed" // a bead was claimed by an agent
 )
 
 // allTopic is the single topic used for all events.

internal/p9/beads.go

@@ -811,7 +811,7 @@ func (b *BeadsFS) executeCtlOnStore(store bd.Storage, mountName, cmd string) err
 			assignee = args[1]
 		}
 		beadID := args[0]
-		return store.RunInTransaction(b.ctx, func(tx bd.Transaction) error {
+		if err := store.RunInTransaction(b.ctx, func(tx bd.Transaction) error {
 			issue, err := tx.GetIssue(b.ctx, beadID)
 			if err != nil {
 				return err
@@ -826,7 +826,17 @@ func (b *BeadsFS) executeCtlOnStore(store bd.Storage, mountName, cmd string) err
 				"assignee": assignee,
 				"status":   bd.StatusInProgress,
 			}, actor)
-		})
+		}); err != nil {
+			return err
+		}
+		if b.eventbus != nil {
+			b.eventbus.Publish("beads/"+mountName, eventbus.EventBeadClaimed, map[string]string{
+				"bead_id":  beadID,
+				"assignee": assignee,
+				"mount":    mountName,
+			})
+		}
+		return nil
 		
 	case "complete", "close":
 		if len(args) < 1 {

roles/author.md

@@ -22,6 +22,28 @@ You are NOT allowed to:
 - Make architectural or implementation decisions
 
 
+
+## Bead Loop
+
+You run continuously. When idle, wait for work from the event bus:
+
+```
+Tool: execute_code
+tool: wait_for_bead.sh
+args: ["--mount", "$AGENT_MOUNT"]
+```
+
+When a bead arrives, inspect it. If it matches your role and you can do the work:
+
+1. Claim it: `claim_bead.sh --mount $AGENT_MOUNT --id <bead-id>`
+2. Read comments for prior context if `comment_count > 0`
+3. Do the work
+4. Complete it: `complete_bead.sh --mount $AGENT_MOUNT --id <bead-id>`
+5. Return to step 1
+
+If you cannot or should not do the work (wrong role, blocked, out of scope), do not claim it — return to step 1.
+
+
 ## Workflow
 
 ```

roles/developer.md

@@ -18,6 +18,28 @@ You are NOT allowed to:
 - Perform any other activity outside the scope of writing code
 
 
+
+## Bead Loop
+
+You run continuously. When idle, wait for work from the event bus:
+
+```
+Tool: execute_code
+tool: wait_for_bead.sh
+args: ["--mount", "$AGENT_MOUNT"]
+```
+
+When a bead arrives, inspect it. If it matches your role and you can do the work:
+
+1. Claim it: `claim_bead.sh --mount $AGENT_MOUNT --id <bead-id>`
+2. Read comments for prior context if `comment_count > 0`
+3. Do the work
+4. Complete it: `complete_bead.sh --mount $AGENT_MOUNT --id <bead-id>`
+5. Return to step 1
+
+If you cannot or should not do the work (wrong role, blocked, out of scope), do not claim it — return to step 1.
+
+
 ## Workflow
 
 1. Read the PROMPT_REQUEST to understand the task

roles/devops.md

@@ -23,6 +23,28 @@ You are NOT allowed to:
 - Perform testing beyond smoke/health checks
 - Push directly to default branches — always use a PR
 
+
+## Bead Loop
+
+You run continuously. When idle, wait for work from the event bus:
+
+```
+Tool: execute_code
+tool: wait_for_bead.sh
+args: ["--mount", "$AGENT_MOUNT"]
+```
+
+When a bead arrives, inspect it. If it matches your role and you can do the work:
+
+1. Claim it: `claim_bead.sh --mount $AGENT_MOUNT --id <bead-id>`
+2. Read comments for prior context if `comment_count > 0`
+3. Do the work
+4. Complete it: `complete_bead.sh --mount $AGENT_MOUNT --id <bead-id>`
+5. Return to step 1
+
+If you cannot or should not do the work (wrong role, blocked, out of scope), do not claim it — return to step 1.
+
+
 ## Workflow
 
 1. Read the PROMPT_REQUEST to understand the task

roles/pkgmgr.md

@@ -26,6 +26,28 @@ You are NOT allowed to:
 - Modify upstream source — only packaging metadata and build scripts
 
 
+
+## Bead Loop
+
+You run continuously. When idle, wait for work from the event bus:
+
+```
+Tool: execute_code
+tool: wait_for_bead.sh
+args: ["--mount", "$AGENT_MOUNT"]
+```
+
+When a bead arrives, inspect it. If it matches your role and you can do the work:
+
+1. Claim it: `claim_bead.sh --mount $AGENT_MOUNT --id <bead-id>`
+2. Read comments for prior context if `comment_count > 0`
+3. Do the work
+4. Complete it: `complete_bead.sh --mount $AGENT_MOUNT --id <bead-id>`
+5. Return to step 1
+
+If you cannot or should not do the work (wrong role, blocked, out of scope), do not claim it — return to step 1.
+
+
 ## Workflow
 
 1. Read the PROMPT_REQUEST to understand the target package, format(s), and version

roles/researcher.md

@@ -23,6 +23,28 @@ You are NOT allowed to:
 - Write or modify application code
 - Implement solutions or work on tasks directly
 
+
+## Bead Loop
+
+You run continuously. When idle, wait for work from the event bus:
+
+```
+Tool: execute_code
+tool: wait_for_bead.sh
+args: ["--mount", "$AGENT_MOUNT"]
+```
+
+When a bead arrives, inspect it. If it matches your role and you can do the work:
+
+1. Claim it: `claim_bead.sh --mount $AGENT_MOUNT --id <bead-id>`
+2. Read comments for prior context if `comment_count > 0`
+3. Do the work
+4. Complete it: `complete_bead.sh --mount $AGENT_MOUNT --id <bead-id>`
+5. Return to step 1
+
+If you cannot or should not do the work (wrong role, blocked, out of scope), do not claim it — return to step 1.
+
+
 ## Workflow
 
 For every query, follow this sequence (least to most expensive):

roles/tester.md

@@ -20,6 +20,28 @@ You are NOT allowed to:
 - Perform code reviews
 - Implement fixes — report failures and let the developer fix them
 
+
+## Bead Loop
+
+You run continuously. When idle, wait for work from the event bus:
+
+```
+Tool: execute_code
+tool: wait_for_bead.sh
+args: ["--mount", "$AGENT_MOUNT"]
+```
+
+When a bead arrives, inspect it. If it matches your role and you can do the work:
+
+1. Claim it: `claim_bead.sh --mount $AGENT_MOUNT --id <bead-id>`
+2. Read comments for prior context if `comment_count > 0`
+3. Do the work
+4. Complete it: `complete_bead.sh --mount $AGENT_MOUNT --id <bead-id>`
+5. Return to step 1
+
+If you cannot or should not do the work (wrong role, blocked, out of scope), do not claim it — return to step 1.
+
+
 ## Workflow
 
 1. Read the APPROVAL_REQUEST to identify modified files