Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx agentmods add commands/metabase/repro-bot/auto-improvegit clone --depth 1 https://github.com/metabase/repro-botWhat it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5 | $0.00000 | $0.00653 |
| Opus 5 | $0.00000 | $0.00327 |
| Sonnet 5 | $0.00000 | $0.00131 |
| Haiku 4.5 | $0.00000 | $0.00065 |
Grade A, and why
auto-improve scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 2d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 56 lines — stays where its author put it; the contents beside it link to each section on GitHub.
/auto-improve - Make Future Repros Better
Analyze past reproduction runs and implement concrete improvements to the repro workflow.
Step 1: Gather Input from SUGGESTIONS.md Files
Glob for repro/*/SUGGESTIONS.md and read all of them. These are the structured output from Phase 8 of each repro run — already distilled and actionable.
Do NOT read raw transcripts (.jsonl). SUGGESTIONS.md already extracts the signal. Only reference transcripts if a specific SUGGESTIONS.md item is unclear and needs more context.
Step 2: Categorize Findings
Sort each suggestion into one of these buckets:
| Category | Files affected | Examples |
|---|---|---|
| MCP server bugs/improvements | mcp/metabase_repro/*.py |
Tool errors, missing params, bad defaults, unhelpful error messages |
| Prompt/command improvements | .claude/commands/repro.md |
Missing steps, unclear instructions, wrong ordering, missing edge cases |
| CLAUDE.md knowledge gaps | CLAUDE.md |
Facts Claude should know but doesn't (e.g., H2 timezone limits, version tag formats, socket REPL vs nREPL) |
| New MCP tools needed | mcp/metabase_repro/*.py, mcp/metabase_repro/server.py |
Functionality gaps that cause fallback to raw Bash commands |
| Permission/workflow issues | Various | Actions that trigger unnecessary user prompts, sandbox violations |
| Memory updates | memory/*.md |
Stable patterns or lessons learned that should persist across sessions |
Step 3: Deduplicate and Prioritize
- Count how many runs hit the same issue (look for similar suggestions across different SUGGESTIONS.md files)
- Estimate token waste per occurrence (e.g., a retry loop wastes ~500 tokens each time; reading a huge file wastes thousands)
- Rank by:
frequency × waste per occurrence - Focus on the top items — don't try to fix everything in one pass
Step 4: Enter Plan Mode
Enter plan mode and present a concrete implementation plan:
- Group changes by file
- Include specific code snippets or text additions (not vague descriptions)
- Include verification steps (how to confirm each change works)
- Call out any changes that might affect existing repro behavior
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 2d ago First seen · 56 lines · 0 tokens per session scan A 55afcade15df
auto-improve is a command published in the GitHub repository metabase/repro-bot (10 stars, last pushed 5mo ago), licensed MIT. It costs nothing until one of its globs matches a file; then it loads 653 tokens. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other commands, from other repositories
checklist
Generate a custom checklist for the current feature based on user requirements.
clarify
Identify underspecified areas in the current feature spec by asking up to 5 highly targeted clarification questions and encoding answers back into the spec.
specify
Create or update the feature specification from a natural language feature description.
analyze
Perform a non-destructive cross-artifact consistency and quality analysis across spec.md, plan.md, and tasks.md after task generation.
constitution
Create or update the project constitution from interactive or provided principle inputs.
converge
Assess the current codebase against the feature's spec, plan, and tasks, then append any remaining unbuilt work as new tasks to tasks.md so implement can complete it.