Agent Run Log
Purpose: Lightweight trace capture for agent runs. Each run becomes a trace that helps us learn process patterns over time.
Related: CONTEXT_GRAPH_APPROACH.md, Agent Registry
How to log a run
When you run an agent or follow a process, add a row to the table below:
- Run ID:
{agent-name}-{date}-{campaign-or-context}(e.g.design-ready-copy-2026-02-04-insurance-broker) - Campaign/Context: Which campaign, client, or use case
- Archetype/Type: If applicable (e.g.
service_2pager,sprint) - Input: File path(s) or inputs used
- Output: File path(s) or outputs generated
- Decisions: Key choices made (why this archetype, which sections, etc.)
- Outcome: Was it used? Did it hit gate? Any edits?
Run log
| Run ID | Campaign/Context | Archetype/Type | Input | Output | Decisions | Outcome | Quality | Time | Date |
|---|---|---|---|---|---|---|---|---|---|
| design-ready-copy-2026-02-04-insurance-broker | insurance-broker-lead-intake | service_2pager | gtm/campaign-launch/campaigns/insurance-broker-lead-intake.md | gtm/marketing-assets/design-ready-copy/insurance-broker-lead-intake-2pager.md | Single service → Service 2-pager (not Sprint, not Seasonal) | Used by designer (Hannah), no major edits | 9/10 | 5 min review | 2026-02-04 |
| ticket-creation-2026-02-05-eden-wikipedia | Eden Wikipedia page data request | linear_ticket | Request via Jonah (PR/Marketing) | clients/eden/linear_tickets/eden-wikipedia-page-data-request.md | Data request → Linear ticket format | Used with edits (title, success criteria, points added) | 8/10 | 0.5 min (vs 10-15 min manual) | 2026-02-05 |
| slack-deployment-2026-02-05-unified-agent-structure | Unified Agent Structure migration | deployment_message | Migration details (12 workers, 3 workflows, shared infrastructure) | gtm/agents/examples/deployments/slack-deployment-unified-agent-structure.md | Non-technical GTM audience focus | Used as-is, shared with team | 8/10 | 1 min (vs 5-10 min manual) | 2026-02-05 |
Pattern notes
As patterns emerge, document them here or in PATTERNS.md:
- [Add patterns as they emerge from multiple runs]