Prompts.
The prompts AGNT uses in production, ready to fork. Every entry has a named author, a last-verified date, and a clear use case. No filler, no "top 10 prompts for X" slop — every prompt ships a real AGNT workflow.
20 prompts · verified 2026-04-10
Memory
3MCP memory-write template
Structured prompt for writing durable facts into AGNT user memory via MCP.
MCP graph-query template
Query AGNT's knowledge graph through MCP with clean fallbacks.
RAG recall guard prompt
Prevent semantic recall from injecting stale or contradictory facts into the agent's prompt.
Debugging
3Gemini CLI scan-lane debugging prompt
Use Gemini CLI's multimodal layer to debug AGNT scan-engine lane failures.
A2A envelope debugging prompt
Trace a failed A2A message through ClawPulse with circuit-breaker awareness.
Cron job doctor prompt
Diagnose and fix stuck APScheduler jobs without stopping the scheduler.
Operations
6Codex fleet-swap operator prompt
Walk an operator through swapping an AGNT agent from claude_local to codex_local.
Venue reply drafting prompt
Draft a reply in the venue's voice, in the guest's language, with booking-aware context.
Heartbeat review prompt
Daily heartbeat review — catch drift between intended and actual agent behavior.
Incident writeup prompt
Turn an incident's raw logs into an AGNT-style post-mortem with timeline + action items.
Daily standup synthesizer
Turn 16 agent heartbeats into a 5-line daily standup for the human lead.
LLM spend anomaly detector
Detect when an agent's spend rate is outside its normal envelope and propose an action.
Content
4Ecosystem release-teardown prompt
The prompt the AGNT signals desk uses to draft ecosystem release breakdowns.
Signal excerpt tightener
Tighten a signal's excerpt to 2 sentences without losing the bridge framing.
Wiki link auditor
Audit an AGNT wiki page for dead links, missing backlinks, and orphan status.
Weekly FAQ compile prompt
Compile the last 7 days of conversations into durable FAQ entries for a venue agent.
Editorial rule
Every prompt earns its slot.
Named author. Last-verified date. Tested-with model. Concrete use case. If a prompt doesn't clear those four bars, it doesn't ship. The library is small on purpose.
Every prompt is something AGNT actually runs in production — the venue intake prompt is the one used during onboarding, the memory-write template is the one wrapped around the MCP tool, the incident writeup is the one we use after a Sev1. No theoretical prompts.
Fork, adapt, ship.
Every prompt is a starting point. Pair it with the right tool in /stack and the right integration in /guides/ecosystem.