Agent Harness Logs

SkillMonitoring & ops

Find, inspect, query, redact, and troubleshoot logs, transcripts, sessions, and OpenTelemetry data from coding-agent harnesses on Windows, macOS, Linux, and WSL. USE FOR: GitHub Copilot in VS Code; Copilot CLI; Claude Code; OpenAI Codex; Gemini CLI; OpenCode; Cursor, Windsurf, Kiro, or other VS Code-family editors; locating log files; diagnosing tool or model failures; correlating turns, subagents, and tool calls; querying JSONL, plaintext, or SQLite without spending model tokens; preparing sanitized diagnostics. DO NOT USE FOR: application logs unrelated to coding agents; uploading raw logs; treating unstable transcript schemas as public APIs.

Available today. Use it from your connected AI after setup.

Connect ahel once, and every AI you use reads what you have installed.

Then ask your AI: use the Agent Harness Logs skill

What this skill tells your AI

The instructions your AI receives, as published by netfabric/netfabric.numerics in .agents/skills/agent-harness-logs/SKILL.md and read by ahel’s review.

Harnesses expose three different evidence surfaces. Identify the surface before querying it.

SurfaceTypical contentBest use
Operational/debug logStartup, extensions, network, retries, crashesWhy the harness failed
Session/transcriptPrompts, responses, tools, turn graphWhat happened in one interaction
OpenTelemetryStructured events, metrics, tracesFleet analysis and durable observability

Workflow

  1. Read references/concepts-and-safety.md; never upload or paste an unreviewed log.
  2. Identify the harness and artifact in references/path-matrix.md.
  3. Load the matching reference: VS Code and editors, Copilot CLI and Claude Code, or Codex, Gemini CLI, and OpenCode.
  4. Run python3 scripts/harness_logs.py discover --toon, then inventory the smallest relevant root with --toon.
  5. Query locally using references/query-recipes.md. Use --toon for model-bound output; stream source files and never read a large log wholesale.
  6. Correlate with stable IDs when available: session, turn, prompt, request, tool-call, span, and parent-span IDs.
  7. Redact to a new file, inspect the result manually, then share only the minimum slice needed.
  8. State harness version, OS, artifact type, time window, and whether a path is documented or discovered.

Script

scripts/harness_logs.py uses the Python standard library for queries and the official @toon-format/cli through npx when --toon is requested. It discovers candidate roots, inventories files, streams regex searches, summarizes JSONL fields, runs read-only SQLite queries, and creates best-effort redacted copies. Structured commands retain JSON/JSONL when --toon is omitted. The script never invokes a model.

References

FileLoad when
references/concepts-and-safety.mdHandling sensitive data, large files, or unstable schemas
references/path-matrix.mdLocating artifacts across operating systems
references/vscode-and-editors.mdTroubleshooting Copilot in VS Code or VS Code-family editors
references/copilot-and-claude.mdInspecting Copilot CLI or Claude Code
references/codex-gemini-opencode.mdInspecting Codex, Gemini CLI, or OpenCode
references/query-recipes.mdRunning deterministic, zero-token queries
references/sources.mdVerifying claims or updating version-sensitive details

Signals

GitHub stars
36
Forks
1
Last commit
Aug 2026
Advanced
Catalog kind
skill
Gateway key
agent-harness-logs
Source
github.com/netfabric/netfabric.numerics