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/emaraschio/cursor-commands/prompt-eval-debuggit clone --depth 1 https://github.com/emaraschio/cursor-commandsWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/commands/emaraschio/cursor-commands/prompt-eval-debug)<a href="https://agentmods.dev/commands/emaraschio/cursor-commands/prompt-eval-debug"><img src="https://agentmods.dev/badge/commands/emaraschio/cursor-commands/prompt-eval-debug.svg" alt="Measured on agentmods" height="20"></a>What 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 | $0.00031 | $0.00664 |
| Opus 5 | $0.00015 | $0.00332 |
| Sonnet 5 | $0.00006 | $0.00133 |
| Haiku 4.5 | $0.00003 | $0.00066 |
Grade A, and why
prompt-eval-debug 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 4d 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 — 45 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Overview
Improve a pasted prompt by designing a tiny eval suite, diagnosing failures, and proposing the smallest next change, without rewriting blindly. Full workflow: .cursor/skill-contracts/prompt-eval-debug/SKILL.md (user install: ~/.cursor/skill-contracts/prompt-eval-debug/SKILL.md).
Defaults
| Setting | Default |
|---|---|
| Eval suite | 1 control (C0) + 3 edge (E1 to E3) + 1 capability-boundary (B1) |
| Prompt change | Smallest next change only (no full rewrite in v1) |
| Scoring | User runs cases manually; skill does not claim CI or automated scores |
| Examples | Generic names only (service-a, task-handler) |
Steps
- Read
.cursor/skill-contracts/prompt-eval-debug/SKILL.mdfor the full agent contract; if that file is missing, read~/.cursor/skill-contracts/prompt-eval-debug/SKILL.md. - Execute phases in order (Intake → Tiny eval suite → Run guidance → Diagnosis → Smallest next change → Deliver).
- Report suite table, diagnoses, and one smallest next change in chat.
Anti-patterns
- Do not rewrite blindly. Trigger: the user asks to rewrite the whole prompt and skip the eval work. Wrong: replacing the entire prompt in one shot with no suite or diagnosis. Correct: produce the tiny eval suite and the smallest next change first, and rewrite fully only after explicit request. Reason: a blind rewrite discards what worked and hides why the prompt failed.
- Keep the capability-boundary case. Trigger: building the five-case suite. Wrong: dropping B1 or expecting the model to succeed without the tools it needs. Correct: include a capability-boundary case that expects escalate, ask, or refuse, and diagnose missing tool or capability when relevant. Reason: instructions are not capabilities, and skipping B1 hides hallucinated success.
- Do not claim catalog CI scored the suite. Trigger: reporting suite outcomes. Wrong: stating automated pass/fail scores the user never ran. Correct: present the suite and ask the user to run it, reporting only results they confirm. Reason: this skill does not execute the catalog CI, so asserting scores would misrepresent what was verified.
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.
- 4d ago First seen · 45 lines · 31 tokens per session scan A e6cd993ee5c1
prompt-eval-debug is a command published in the GitHub repository emaraschio/cursor-commands (9 stars, last pushed 27d ago), licensed MIT. It adds 31 tokens to every session and 664 once invoked, about $0.0002 per session on Opus 5. 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
grok_init
Discover, verify, and persist the local HTTP proxy used by Grok Build Supervisor.
analyze-deps
Analyze the project's dependency tree, identify unused or duplicate dependencies, and suggest optimizations.
check-coverage
Check test coverage, identify gaps, and suggest which code needs tests.
security-audit
Perform a security-focused review of the code I've @-mentioned (or the most relevant changes in this conversation).
fix-issue
Command "fix-issue" from HKTITAN/cursor-best-practices, covering steps and rules.
run-tests-and-fix
Run the project's test suite, fix any failures, and report what changed.