- Home
- Skills
- Agents & Orchestration
- Freebies — Agent Log Parser
Works with the AI Agents you already use
Freebies — Agent Log Parser
Parse agent logs to detect circular reasoning loops, tool latency bottlenecks, and token consumption metrics.
Free
Freebies — Agent Log Parser
Example session with this skill installed
Analyze the agent logs in ./logs/execution-trace.jsonl. I need to know why the file_search tool is taking so long and if there are any repeating loops.
- Read your context and instructions
- Compiled the freebies
I have parsed the execution trace from ./logs/execution-trace.jsonl. The analysis identified a specific bottleneck in the file_search tool and one circular reasoning loop during the indexing phase.
Agent Execution Analysis Report
| Metric | Value | Status |
|---|---|---|
| Total Tool Calls | 42 | Nominal |
| Total Token Usage | 18,450 | Within Budget |
| Circular Loops Detected | 1 | Critical |
| Avg. Tool Latency | 1.8s | Nominal |
| Max Tool Latency | 12.4s (file_search) | Warning |
Latency Distribution
read_file: 150ms (avg)list_dir: 210ms (avg)file_search: 12.4s (max) - Bottleneck detected during deep recursive globbing.
Circular Reasoning Detection
Found 1 loop at Step 14-18. The agent attempted to call edit_file on config.json four times with identical original_text blocks, receiving a "no changes detected" error each time without updating its reasoning strategy.
Token Utilization
- Input Tokens: 12,200
- Output Tokens: 6,250
- Step 12 (Reasoning): Highest single-step consumption (2,400 tokens).
Next steps
- Review the
file_searchimplementation to limit depth or excludenode_modules. - Update the system prompt to handle "no changes detected" errors to break the
edit_fileloop. - Check
config.jsonpermissions to ensure the agent has write access.
Connects securely to your tools. The creator never sees your data.
About this skill
Debugging autonomous agents is difficult when execution traces are thousands of lines long. This skill parses agent logs to pinpoint exactly where tool calls fail, why reasoning loops occur, and how much each step costs in tokens.
What it does
- Log Parsing reads and structures JSONL and OpenTelemetry traces from agent runs.
- Loop Detection identifies circular reasoning where agents repeat failing tool calls with identical arguments.
- Performance Analytics calculates tool latency distributions and execution time per step.
- Cost Tracking measures token utilization across the entire execution sequence.
- Optimization Reporting generates summaries of failed calls and dead-end iterations.
How it works
- Ingest Logs by pointing the skill to your local execution trace files or JSONL logs.
- Analyze Traces to extract tool call sequences, timestamps, and reasoning patterns.
- Scan for Faults including circular logic, high-latency bottlenecks, and token spikes.
- Generate Report containing a structured breakdown of agent performance and failure points.
Frameworks & tools
Built for the AAIF agentskills.io Open Standard. Compatible with Claude Code, Agent SDK v0.3+, Gemini CLI, Cursor, Windsurf, and OpenClaw. Works with JSONL and OpenTelemetry formatted logs.
Why this beats prompting it yourself
Generic prompts often miss subtle circular reasoning loops or fail to calculate precise latency distributions across large log files. This skill uses a deterministic workflow designed specifically for the Domain 06 MCP agents swarm scope to ensure no execution bottleneck is overlooked.
Use cases
- Debugging agent "hallucination loops" where the model gets stuck in a tool-call retry cycle.
- Identifying slow MCP server tools that are bottlenecking multi-agent swarm performance.
- Auditing token consumption to find expensive reasoning steps that need prompt optimization.
- Post-mortem analysis of failed autonomous task runs to prevent future regressions.
Known limitations
Confined to the local project workspace. Does not support real-time streaming log analysis or remote cloud logging providers directly.
Find more skills like this on the Claude Skills Marketplace, the Codex Skills Marketplace and the Gemini Skills Marketplace, or browse Claude Skills for Agents & Orchestration.
Restricted to local workspace files. No support for remote log streaming or direct cloud provider integration. Requires logs in JSONL or OpenTelemetry format.
How to use
- 1
Connect your agent to the Agensi MCP
Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup
- 2
Sign in to Agensi
Your agent shows a sign-in link. Sign in and click Allow.
- 3
Ask for what you need
Your agent loads this skill and uses it on your task. It always asks before buying a skill.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill goes through the trust checks below.
Security scanned
Checked by our 8-point scan before listing
Fresh listing
Recently published to Agensi
Free to use
No account required to browse
Trust & safety
Security scan
Security scan passed 1 day ago
- Free with an Agensi account
Linux only
- Software
- Zero global installationorelevated privileges required.
File System Read/Write (strictly scoped to current project workspace `./`).
Permissions
Target Use Cases: - Analyze this AI agent execution log for circular loops. - Parse agent tool call traces and identify token waste. - Diagnose why an autonomous agent got stuck in a loop.