Dead-letter triage over MCP: investigate → explain → replay Added in 1.6.0
Investigate and act on dead letters across a CritterWatch-monitored fleet through the MCP dead-letter tools: summarize_dead_letters (grouped triage counts by message/exception type), query_dead_letters (individual envelopes with the ids replay/discard need), then replay_dead_letters / discard_dead_letters.
Use when an operator or agent asks 'what's failing and why', 'what's in the dead letter queue', 'explain these failures', or 'replay the recoverable ones and discard the rest'. Covers the partial-answer discipline (databasesAnnounced vs databasesAnswered — never report an empty queue on a partial read), capability gating (mcp.dlq.read vs dlq.replay/dlq.discard), and seeding test dead letters.
Skill details
- Skill name:
wolverine-integrations-critterwatch-dlq-triage - Applies to:
wolverine - Tags:
observability,critterwatch,mcp,dead-letters,dlq,triage,operations,agents - Added in: 1.6.0
- License: Proprietary. Copyright JasperFx Software LLC.