recording-what-an-agent-did

Record agent-run tool calls, edits, inputs, outputs, tokens, and approvals as auditable trace evidence.

33|Updated May 24, 2026
One-click install
npx skills add https://github.com/FlyFission/nuclear-grade-context-engineering --skill recording-what-an-agent-did
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: recording-what-an-agent-did
Source: https://github.com/FlyFission/nuclear-grade-context-engineering/tree/main/skills/recording-what-an-agent-did
Command: npx skills add https://github.com/FlyFission/nuclear-grade-context-engineering --skill recording-what-an-agent-did

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Captures a clear, repeatable record of an agent run, including tool calls, decisions, inputs, outputs, token use, and approval steps, so debugging, auditing, cost review, and release decisions are reproducible and defensible.

Core Features & Use Cases

  • Per-step trace records: captures the inputs, outputs, results, and evidence status for tool calls, edits, commands, and API calls.
  • Trace linkage: connects run evidence to the packet's trace.md and verification.md for end-to-end auditability.
  • Visibility and governance: summarizes token usage, latency, and approvals to support cost reviews and release decisions.
  • Escalation-ready: supports documenting gaps, failures, and recovery steps for post-incident analysis.

Quick Start

Trace an agent run and generate a structured run-evidence record linked to the packet's trace and verification data.

Frequently Asked Questions about recording-what-an-agent-did

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I record agent runs for debugging and auditing?

Recording agent runs for auditing captures per-step trace records of tool calls, inputs, outputs, and approvals, linking them to verification data so debugging and release decisions are reproducible and defensible.

What is agent run trace verification?

Agent run trace verification connects recorded evidence like tool calls, token usage, and approvals to a packet's trace and verification files, ensuring end-to-end auditability for AI workflows.

Does this trace recording capture token usage and cost data?

Trace recording captures token usage, latency, and approval steps to support cost reviews and governance, summarizing run evidence for incident analysis and release decisions.

How do I document agent failures for post-incident analysis?

Documenting agent failures for post-incident analysis relies on escalation-ready trace records that capture gaps, failures, and recovery steps alongside per-step tool call evidence.

Can I export agent run summaries for cost review?

Exporting agent run summaries for cost review provides structured trace rows and run summaries that link tool calls, edits, and approvals directly to release decisions.

What's the best way to automate evidence recording for packet-based AI workflows?

Automating evidence recording for packet-based AI workflows captures per-step inputs, outputs, and decision points with applied limits, exporting structured trace rows for audits.