docs(atlas): add agent audit event template

2026-05-25 18:16:36 -07:00
parent 9747d26a18
commit da24b029b3
2 changed files with 51 additions and 0 deletions
@@ -23,6 +23,7 @@ Canonical templates for durable Atlas outputs. Use these when chat, Kanban, or s
 - [[test-report]] — test, benchmark, evaluation, and smoke-check reports.
 - [[status-report]] — weekly/monthly/milestone status updates.
 - [[postmortem]] — incident/failure writeups.
+- [[agent-audit-event]] — short audit notes for agent actions: model, provider, tools/MCP, data touched, outcome, cost, and latency.
 - [[reviewer-checklist]] — common review checklist for all artifacts.

 ## Destination policy
@@ -46,6 +47,7 @@ Use lowercase kebab-case filenames with a date prefix where chronology matters.
 - Test reports: `YYYY-MM-DD-<scope>-test-report.md`
 - Status reports: `YYYY-MM-DD-status-report.md`
 - Postmortems: `YYYY-MM-DD-<incident-name>-postmortem.md`
+- Agent audit events: `YYYY-MM-DD-<agent-or-system>-<event-topic>-audit.md`

 ## Promotion workflow

@@ -0,0 +1,49 @@
+---
+type: agent-audit-event
+project: "{{project}}"
+status: draft
+owner: "Atlas"
+created: {{date}}
+updated: {{date}}
+source: "{{source-session-or-kanban-task}}"
+tags: [atlas, audit, agent-event]
+---
+
+# Agent Audit Event — {{event-title}}
+
+Use this short note when an Atlas/Hermes agent action should be auditable but does not need a full status report or postmortem.
+
+## Summary
+
+- **When:** {{timestamp-or-date}}
+- **Actor:** {{atlas/profile/agent-name}}
+- **Trigger:** {{user-request | cron-job | webhook | kanban-task | manual-ops}}
+- **Outcome:** {{success | partial | failed | cancelled}}
+- **One-line result:** {{what changed or what was learned}}
+
+## Runtime
+
+- **Model:** {{model-name}}
+- **Provider:** {{provider-name}}
+- **Profile/session:** {{profile-name-or-session-id}}
+- **Tools used:** {{terminal, file, web, browser, etc.}}
+- **MCP servers / external services:** {{server-or-service names, or none}}
+
+## Data touched
+
+- **Local paths:** {{paths read/written, or none}}
+- **Remote systems:** {{GitHub, Gitea, Google, n8n, Obsidian API, etc., or none}}
+- **Secrets/PII exposure:** {{none | brief description + mitigation}}
+- **Writes/side effects:** {{files changed, messages sent, jobs scheduled, services restarted, or none}}
+
+## Cost and latency
+
+- **Approx. cost:** {{amount, token cost, or unknown/not reported}}
+- **Latency / runtime:** {{duration, or unknown/not reported}}
+- **Notable performance notes:** {{slow tool, retry, timeout, cache hit/miss, or none}}
+
+## Evidence
+
+- **Verification:** {{test/check/log/URL/commit/path proving outcome}}
+- **Artifacts:** {{links to reports, diffs, generated files, or none}}
+- **Follow-up:** {{next action, owner, or none}}