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
- Prefer the installed
agenttrace binary when it is available on PATH.
- If the binary is not available and the current directory is the
luoyuctl/agenttrace repository, use cargo run -q -p agenttrace --.
- Start with discovery unless the user gave a specific file or directory:
agenttrace --doctor
agenttrace --overview
- For a fast human report, use Markdown:
agenttrace --overview -f markdown -o agenttrace-overview.md
- For automation or CI, use JSON or health gates:
agenttrace --overview -f json -o agenttrace-overview.json
agenttrace --overview --fail-under-health 80 --fail-on-critical --max-tool-fail-rate 15
- For a single recent session:
agenttrace --latest
agenttrace --latest -f json
- For a specific export or session directory:
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.
1---2name: agenttrace-session-audit3description: Audit local AI coding-agent sessions with agenttrace for cost, tokens, tool failures, latency, anomalies, health, diffs, and CI gates.4license: MIT5---6
7# agenttrace Session Audit
8
9Use 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.
10
11## Workflow
12
131. Prefer the installed `agenttrace` binary when it is available on `PATH`.
142. If the binary is not available and the current directory is the `luoyuctl/agenttrace` repository, use `cargo run -q -p agenttrace --`.
153. Start with discovery unless the user gave a specific file or directory:
16
17```bash
18agenttrace --doctor
19agenttrace --overview
20```
21
224. For a fast human report, use Markdown:
23
24```bash
25agenttrace --overview -f markdown -o agenttrace-overview.md
26```
27
285. For automation or CI, use JSON or health gates:
29
30```bash
31agenttrace --overview -f json -o agenttrace-overview.json
32agenttrace --overview --fail-under-health 80 --fail-on-critical --max-tool-fail-rate 15
33```
34
356. For a single recent session:
36
37```bash
38agenttrace --latest
39agenttrace --latest -f json
40```
41
427. For a specific export or session directory:
43
44```bash
45agenttrace path/to/session-or-export.json
46agenttrace --overview -d path/to/session-dir
47```
48
49## Report Focus
50
51- Lead with the highest-risk sessions and the reason they matter.
52- Call out token/cost waste, repeated tool failures, retry loops, long gaps, and low health scores.
53- When proposing a CI gate, include the exact `agenttrace` command and threshold.
54- If no sessions are detected, run `agenttrace --doctor` and report the detected agent directories and next step.
55- Report the session capability level (`Detailed`, `Aggregate`, or `Limited`) before relying on latency or step evidence.
56- Treat Tool Steps as metadata-only evidence. Do not imply that Aggregate or Limited sources have a complete execution trace.
57
58## Guardrails
59
60- Treat prompts, code, and session contents as local/private data. Do not upload logs to external services.
61- Do not invent metrics. If a parser cannot infer cost, model, or latency, say which field is missing.
62- Do not compare missing event-level evidence as zero latency or zero failures; mark it unavailable.
63- Do not overwrite user reports unless the user asked for that output path.
gabrielmoreira