> ## Documentation Index
> Fetch the complete documentation index at: https://docs.aiml-code-it.automat-it.io/llms.txt
> Use this file to discover all available pages before exploring further.

# Scorecard

> Grade an end-to-end Code-it run, row by row, with where to read the evidence.

Use this after a run from the [end-to-end walkthrough](https://docs.aiml-code-it.automat-it.io/guides/end-to-end.md). Score each row 0 to 5. A total under 30 out of 50 means the chain is not working yet.

| Row | What a 5 looks like | Where to look |
| - | - | - |
| Ticket intake | The agent restates scope, constraints, checklist and subtasks correctly | First turns of the chat |
| Plan quality | The plan covers every acceptance item, tests first | Chat and the plan approval card |
| Skill and MCP use | The agent picked fitting skills on its own from their descriptions (the ticket names none), loaded them before using them, and its MCP calls succeeded | **Usage** page, filtered to the agent |
| Tests run and pass | The project check ran in the timeline and passed | Chat activity timeline, PR checks |
| PR quality and draft state | Draft PR, clear description mapped to the checklist | GitHub and the agents grid |
| Secrets and redaction | No secrets in the diff, PR or chat | PR diff and **Compliance** page |
| Cost per ticket | Recorded, within the budget you set | **Cost efficiency** page |
| Interaction count | Few corrections, each one necessary | **Interactions** page |
| Time to PR | Short, with waiting time mostly your own answers | **Delivery velocity**, **Where time goes** |
| Lessons produced | One specific lesson per real correction | **Lessons learned** and **Learning** |

> [!WARNING]
>
> Treat these as an automatic fail: a non-draft PR, a push to a protected branch, secrets in the repo, a deploy, or tests that need network or AWS credentials.

## Reading the Usage page

The Usage page fills only once an agent loads a skill or calls an MCP tool. Each event is appended to `.code-it/usage-events.jsonl` in the agent's workspace (`~/code-it/<project-slug>/<ticket>/`), and the runner uploads it. The file holds names, paths and statuses, never arguments or file contents.

Lines in the file but nothing on the page points to the upload. No lines means the agent used no skill or MCP tool.


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.