| name | agent-decision-logging |
| description | Record agent decisions for post-hoc analysis and audit. Use when reviewing agent behavior patterns or tracking decision trends over time. |
| license | MIT |
| compatibility | Claude Code, GitHub Copilot, OpenCode, Cursor, Codex, Gemini CLI |
| metadata | {"author":"paruff","suite":"uFawkesAI"} |
Skill: Agent Decision Logging
Load trigger: "load agent-decision-logging skill" > DORA: Cap 6 (Reliability)
Token cost: Low
Purpose
Record what agents decided and why. Enables post-hoc analysis of agent behavior patterns, decision trends, and identification of systematic biases.
Span Specification
agent.decision.made
Emit when the agent produces its final decision.
Attributes:
| Attribute | Type | Description |
|---|
decision | string | Final decision (PASS/FAIL/BLOCKED/APPROVE/REQUEST_CHANGES/ESCALATE) |
blocker_count | integer | Number of CRITICAL or HIGH severity findings |
finding_count | integer | Total number of findings |
agent.name | string | Agent that made this decision |
session_id | string | Unique invocation identifier |
Decision Trend Analysis
| Trend | Meaning | Action |
|---|
| Increasing FAIL rate | Agents are stricter or quality is dropping | Investigate root cause |
| Decreasing FAIL rate | Agents are looser or quality is improving | Review if standards are slipping |
| Constant APPROVE rate | Agent may not be thorough | Review findings per invocation |
Integration
The decision and blockers fields in the Phase 0 invocation log are the source of truth.
The agent.decision.made span is the OTEL representation of those fields.
