fix(protocol-skill): fix prose inconsistencies after #496 lifecycle reorder

Fix Ready Messages intro paragraph to say agents announce presence when
joining (before init), not after completing initialization. Fix Common
Mistakes 'Post intent before others are ready' row to reference
init-complete info message, not just ready message.

AI-Used: [claude]

Author: toejough

skills/use-engram-chat-as/SKILL.md

@@ -469,7 +469,7 @@ Context: Discovered during build failure -- running go test directly misses cove
 
 ## Ready Messages
 
-Agents announce they have completed initialization and are watching the chat. This provides a synchronization point.
+Agents announce their presence when joining the chat. This provides an early synchronization point before initialization completes.
 
 ```toml
 [[message]]
@@ -700,7 +700,7 @@ tail -n +$((CURSOR + 1)) "$CHAT_FILE"
 | Skip `ready` message or post it late | Post `ready` as your FIRST action — before reading history or loading resources. Presence before initialization, not after. |
 | Emit `learned` for trivial observations | Only emit when knowledge is reusable across sessions |
 | Ignore `shutdown` message | Exit monitor Agent loop after completing in-flight work and posting `done` |
-| Post intent before others are ready | In lead setup: wait for expected `ready` messages (30s timeout) |
+| Post intent before others are ready | In lead setup: wait for the agent's init-complete `info` message (30s timeout from that message, not from `ready`) |
 | Use old field names (`[[entry]]`, `message =`) | Clean break: use `[[message]]` and `text =` |
 | Truncate or delete the chat file | Chat files are persistent -- prior sessions contain valuable context |
 | Skip heartbeat in long-lived reactive agent | Post heartbeat every 5 min so others know you're alive |