tt-a1i

Agent Insight — Development skill for Claude Code

Development community

Local-first, cross-agent session insights for Claude Code, Codex, Cursor, OpenCode, and Pi.

How to install Agent Insight

This entry records only its repository, not the path inside it, so there is no exact command to give. Open tt-a1i/agent-insight and copy the folder into ~/.claude/skills/, or the file into ~/.claude/agents/.

What Agent Insight does

Local-first, cross-agent session insights for Claude Code, Codex, Cursor, OpenCode, and Pi.

Alternatives in Development

IPolloWork — Enterprise-grade, local-first Agent Workbench for people and agent teams 4.9k ★
Ralphy — My Ralph Wiggum setup, an autonomous bash script that runs Claude Code, Codex, OpenCode, Cursor agent, Qwen & 2.8k ★
OpenSquirrel — For people who get distracted by agents 1.3k ★

README

Agent Insight

**One fused coding-agent report: Claude-compatible Insights plus a sharp, evidence-backed audit of how you work with agents.**

Agent Insight analyzes local session history from Claude Code, Codex, Cursor, OpenCode, and Pi.

Semantic work stays with the model already selected in the invoking host. Agent Insight never silently switches providers or starts another agent CLI.

The project has two goals:

Implement a Claude-compatible /insights baseline as a fixture-protected compatibility profile (observable Claude Code 2.1.206 behavior as design reference — not an official Claude certification claim).
Extend that baseline across agents: Host/Source decoupling, explicit time ranges, honest coverage, resumable execution, and a user-audit extension.

**Product model:** any available Host + that Host's current Model + any selected local Source corpus → one fused Agent Insight report.

Feature implementation is complete under that model. Agent Insight does **not** claim official Claude parity certification.

The shared command is `/agent-insights`. Codex exposes the same workflow as `$agent-insights`. Agent Insight does not replace Claude Code's built-in `/insights` command.

**Default report language is Simplified Chinese** (`--locale zh`): HTML chrome (section titles, empty states, audit headings) and model-generated prose. The product title stays English `Agent Insight`. Pass `--locale en` for English chrome and prose. On Claude host, the HTML shell remains English `Claude Code Insights` for the compatibility profile regardless of locale; use `--locale en` when you also want English model prose.

The compatibility profile and every intentional extension are defined in the [Claude Code 2.1.206 compatibility contract](docs/parity/claude-2.1.206.md).

Why Agent Insight

One report, not two tools: Claude-compatible baseline sections and the user audit are generated in one run.
One workflow across agents: analyze the current agent, s