SKILL.md
---
name: agenttrace-session-audit
description: Audit local AI coding-agent sessions with agenttrace for cost, tokens, tool failures, latency, anomalies, health, diffs, and CI gates.
category: observability
license: MIT
requires:
bins: [agenttrace]
---
# agenttrace Session Audit
Use this skill when session logs need an operational read: spend, token burn, cache use, tool failures, retry loops, latency, health, anomalies, and CI gate readiness.
## Workflow
1. Prefer the installed `agenttrace` binary when it is available on `PATH`.
2. If the binary is not available and the current directory is the `luoyuctl/agenttrace` repository, use `cargo run -q -p agenttrace --`.
3. Start with discovery unless the user gave a specific file or directory:
```bash
agenttrace --doctor
agenttrace --overview
```
4. For a fast human report, use Markdown:
```bash
agenttrace --overview -f markdown -o agenttrace-overview.md
```
5. For automation or CI, use JSON or health gates:
```bash
agenttrace --overview -f json -o agenttrace-overview.json
agenttrace --overview --fail-under-health 80 --fail-on-critical --max-tool-fail-rate 15
```
6. For a single recent session:
```bash
agenttrace --latest
agenttrace --latest -f json
```
7. For a specific export or session directory:
```bash
agenttrace path/to/session-or-export.json
agenttrace --overview -d path/to/session-dir
```
## Report Focus
- Lead with the highest-risk sessions and the reason they matter.
- Call out token/cost waste, repeated tool failures, retry loops, long gaps, and low health scores.
- When proposing a CI gate, include the exact `agenttrace` command and threshold.
- If no sessions are detected, run `agenttrace --doctor` and report the detected agent directories and next step.
- Report the session capability level (`Detailed`, `Aggregate`, or `Limited`) before relying on latency or step evidence.
- Treat Tool Steps as metadata-only evidence. Do not imply that Aggregate or Limited sources have a complete execution trace.
## Guardrails
- Treat prompts, code, and session contents as local/private data. Do not upload logs to external services.
- Do not invent metrics. If a parser cannot infer cost, model, or latency, say which field is missing.
- Do not compare missing event-level evidence as zero latency or zero failures; mark it unavailable.
- Do not overwrite user reports unless the user asked for that output path.