mcpField
C
ccbox

ccbox

github.com/diskd-aiAgent workflows39121 installsUpdated 35w ago
Repository

About ccbox

ccbox is a CLI for inspecting local agent session logs. It helps users review projects and sessions, then assemble evidence-based reports from recorded actions and outputs.

Key features

  1. 01Discover local projects
  2. 02List project sessions
  3. 03Inspect full session logs
  4. 04Review tool calls and outputs
  5. 05Prepare insights reports

Use cases

  1. 01Reviewing the latest session for a project
  2. 02Summarizing an agent's actions
  3. 03Finding errors and risks in logs

Install

Copy the command and run it in a terminal.

$npx skills add https://github.com/diskd-ai/ccbox --skill ccbox

After installing, type/ccboxin Claude Code to confirm it loaded.

Bundled files

References and scripts installed along with SKILL.md.

README.md
agents/
openai.yaml
references/
cli.md
insights.md

Details

From the project README

ccbox

Inspect local agent session logs via ccbox CLI and produce quick, evidence-based insights.

Quick start

  • List discovered projects:
ccbox projects
  • List sessions for the current folder (auto-detect project):
ccbox sessions
ccbox sessions --limit 50 --offset 0 --size
  • Inspect the latest session for the current folder:
ccbox history
ccbox history --full
ccbox history --limit 200 --offset 0 --full --size
  • Inspect a specific project or log file:
ccbox sessions "/abs/path/to/project"
ccbox history "/abs/path/to/project" --full
ccbox history "/abs/path/to/session.jsonl" --full

Workflow (recommended)

  1. Confirm ccbox is available and up to date:
    • ccbox --help
    • ccbox update (prints up-to-date: or updated:; self-update is supported on macOS/Linux)
  2. Gather facts first (do not guess):
    • Use ccbox sessions to select the latest session for the current folder, or ccbox sessions "/abs/project/path" for an explicit project.
    • Copy the log_path from the first line.
    • Run ccbox history "/abs/log.jsonl" --full to capture evidence (tool calls, tool outputs, errors).
  3. Produce an insights report:
    • What the user wanted (first USER prompt).
    • What the agent did (sequence of tool calls + key outputs).
    • What changed (files/commands visible in tool outputs).
    • What failed or is risky (warnings, truncated timelines, panics/errors in outputs).
    • Next steps (concrete verification commands and follow-ups).

References

  • CLI commands + output formats: references/cli.md
  • Insights checklist + report template: references/insights.md

Comments

Markdown supported