S
pending
My AI Coding Agent a Performance Review
Grounded / Real
Inflated / Uruttu
Original Content
I gave my AI coding agent a formal performance review. It got an A-. I got a B.
Claude Code saves a transcript of every session on your machine. 31 days of mine, sitting in a folder. So I built skiplevel, an open source tool that reads them and reviews both parties.
What the review found:
The agent said "you're absolutely right" 56 times. It read the same file 30 times in one session. It also generated an itemized receipt showing $164 of usage went to re-reading files it had already read.
My side was not clean either. 339 interruptions. 3,025 words typed in all caps. I apologized to a language model 7 times, which it noted in my file.
Everything runs locally in about 2 seconds, nothing uploads, and it works on Claude Code, Codex CLI, and opencode. The grading rubric is deterministic and documented, no LLM involved.
Built the whole thing with Claude Code over a weekend. So Claude wrote the tool that reviews Claude, then graded itself A-.
Free and open source, link in comments. Run it and tell me your verdict
Claude Code saves a transcript of every session on your machine. 31 days of mine, sitting in a folder. So I built skiplevel, an open source tool that reads them and reviews both parties.
What the review found:
The agent said "you're absolutely right" 56 times. It read the same file 30 times in one session. It also generated an itemized receipt showing $164 of usage went to re-reading files it had already read.
My side was not clean either. 339 interruptions. 3,025 words typed in all caps. I apologized to a language model 7 times, which it noted in my file.
Everything runs locally in about 2 seconds, nothing uploads, and it works on Claude Code, Codex CLI, and opencode. The grading rubric is deterministic and documented, no LLM involved.
Built the whole thing with Claude Code over a weekend. So Claude wrote the tool that reviews Claude, then graded itself A-.
Free and open source, link in comments. Run it and tell me your verdict
Validated Content
What checks out as accurate/plausible:
- Claude Code does save local session transcripts — this is genuinely how Claude Code works; sessions are logged in JSON/JSONL format on your machine, and multiple real open-source tools exist specifically to parse these transcripts (e.g.,
claude-code-transcripts, cost/diagnostic parsers that report tool-call distribution and expensive turns — this is an established category of tooling) - "You're absolutely right" as a recurring Claude Code phrase — this is a widely-documented, commonly-joked-about verbal tic of Claude Code in the developer community, so this detail rings true
- Deterministic, non-LLM grading on local transcripts — technically plausible; parsing tool-call counts, repeated file reads, and cost estimates from a transcript doesn't require an LLM at all, just log analysis
- Multi-tool support (Claude Code, Codex CLI, opencode) — plausible, since these tools share broadly similar session-log structures
Couldn't independently verify:
- I was unable to find a specific open-source project named "skiplevel" matching this description in current search results. This doesn't mean it doesn't exist — it may be too new, unindexed, or the link is only "in comments" as stated — but I can't confirm the tool itself, its GitHub repo, or the specific numbers (56 mentions, $164, 339 interruptions) independently.