AgentEval tracing.instructions.md

A set of guidelines for recording and replaying AI-agent activity. A trace is a saved record of prompts, responses, workflow steps, timing, and related information.

In plain words
What is it for?
It helps record conversations and multi-agent workflows, save traces as JSON, and replay agent or workflow traces deterministically.
Why use it?
It makes agent behavior easier to inspect and reproduce instead of relying on a live run that may produce different results each time.

Instructions file for GitHub Copilot

Install

Getting it into your agent

One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.

agentmods
npx agentmods add instructions/agentevalhq/agenteval/tracing
Clone the repo
git clone --depth 1 https://github.com/AgentEvalHQ/AgentEval

Made for: GitHub Copilot.

Per session 999 This file is loaded in full into every session.
When invoked 999 The same file — it is already loaded in full.
Security scan A 0 findings. Scan, not verified.
Origin original No closer match found in the catalogue.
Token cost

What it costs to keep this loaded

Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.

ModelPer sessionOnce invoked
Fable 5 $0.00999 $0.00999
Opus 5 $0.00500 $0.00500
Sonnet 5 $0.00200 $0.00200
Haiku 4.5 $0.00100 $0.00100

Measured yesterday against content hash 4ce918262e69, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

AgentEval tracing.instructions.md scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured yesterday.

A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.

Nothing flagged

None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.

.github/instructions/tracing.instructions.md · 170 lines

How it starts

The opening of the file, as written. The whole thing — 170 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Tracing Implementation Guidelines

Core Tracing Components

Recording Agents

  • TraceRecordingAgent - Wraps agent to capture executions
  • ChatTraceRecorder - Records multi-turn conversations
  • WorkflowTraceRecorder - Records multi-agent workflow steps

Replay Agents

  • TraceReplayingAgent - Replays recorded traces deterministically
  • WorkflowTraceReplayingAgent - Replays workflow traces

Serialization

  • TraceSerializer - Save/load AgentTrace to/from JSON
  • WorkflowTraceSerializer - Save/load WorkflowTrace to/from JSON

AgentTrace Structure

public class AgentTrace
{
    public string Version { get; set; }
    public string TraceName { get; set; }
    public DateTimeOffset CapturedAt { get; set; }
    public string? AgentName { get; set; }
    public string? ModelId { get; set; }
    public List<TraceEntry> Entries { get; set; }
    public TracePerformance? Performance { get; set; }
    public Dictionary<string, object>? Metadata { get; set; }
}

public class TraceEntry
{
    public TraceEntryType Type { get; set; }
    public int Index { get; set; }
    public string? Prompt { get; set; }
    public string? Text { get; set; }
    public long? DurationMs { get; set; }
    public TraceTokenUsage? TokenUsage { get; set; }
    public List<TraceToolCall>? ToolCalls { get; set; }
    public TraceError? Error { get; set; }
    public bool IsStreaming { get; set; }
    public List<TraceStreamChunk>? StreamingChunks { get; set; }
}

Recording Pattern

// Wrap real agent
await using var recorder = new TraceRecordingAgent(realAgent, "weather_test");

// Execute (calls real agent, captures result)
var response = await recorder.InvokeAsync("query");

// Get trace for storage
var trace = recorder.Trace;

// Save to file
await TraceSerializer.SaveToFileAsync(trace, "trace.json");

Replay Pattern

// Load saved trace
var trace = await TraceSerializer.LoadFromFileAsync("trace.json");

// Create replayer
var replayer = new TraceReplayingAgent(trace);

// Replay entries in order
while (!replayer.IsComplete)
{
    var response = await replayer.InvokeAsync("prompt");
    // Response is identical to original
}

Read the full file on GitHub · 170 lines

Changes

What this file has done since we first saw it

Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.

  1. yesterday First seen · 170 lines · 999 tokens per session scan A 4ce918262e69

Subscribe to this mod's changes

AgentEval tracing.instructions.md is an instructions file published in the GitHub repository AgentEvalHQ/AgentEval (138 stars, last pushed yesterday), licensed MIT. It adds 999 tokens to every session, about $0.0050 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.