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 skills add ngocsangyem/MeowKit --skill prompt-enhancergit clone --depth 1 https://github.com/ngocsangyem/MeowKitWrote 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/skills/ngocsangyem/meowkit/prompt-enhancer)<a href="https://agentmods.dev/skills/ngocsangyem/meowkit/prompt-enhancer"><img src="https://agentmods.dev/badge/skills/ngocsangyem/meowkit/prompt-enhancer/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/ngocsangyem/meowkit/prompt-enhancer"><img src="https://agentmods.dev/badge/skills/ngocsangyem/meowkit/prompt-enhancer.svg" alt="Reviewed on agentmods" width="80" 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 5.1 | $0.00105 | $0.03422 |
| Opus 5 | $0.00053 | $0.01711 |
| Sonnet 5 | $0.00021 | $0.00684 |
| Haiku 4.5 | $0.00011 | $0.00342 |
Grade A, and why
mk:prompt-enhancer 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 8d 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 — 214 lines — stays where its author put it; the contents beside it link to each section on GitHub.
mk:prompt-enhancer
Refine a draft user prompt using a 5-step framework — mode-select, decompose,
detect, (scout if --deep), map, rewrite — grounded in 7 source docs
(Anthropic / OpenAI Codex / FactoryAI prompting + Anthropic context-engineering
- skill-authoring rules). The rewrite uses the universal kernel only (the rule is Hard Constraints item 4).
When to use
- Draft prompt is ambiguous, missing context, weakly structured, or coupled to one model's framing (XML, role-as-XML, "think step by step", etc.).
- User says: "enhance prompt", "optimize prompt", "make this prompt better", "rewrite this prompt".
- Want to see WHY a prompt was rewritten: append
--analyze. - Want a numeric quality score on the original prompt: append
--score(auto-promotes to--analyze --score). - For codebase-aware suggestions: append
--deep(opt-in only). - Explicit:
/mk:prompt-enhancer [--analyze] [--score] [--deep] [--save-to <path>] [text]. - Architecture-review prompts: use the recipe in
references/architecture-review-mode.mdvia--analyze --deep— it rewrites the prompt to ASK for a review; it never performs one (that ismk:review). Recipe is opt-in via those flags; default-mode output is unchanged.
When NOT to use
- Generating a prompt from scratch →
mk:brainstorming(technical) /mk:office-hours(product). - Re-examining a code review or plan →
mk:elicit. - Building an implementation plan →
mk:plan-creator. - Auditing running output of an LLM →
mk:evaluateormk:review. - General codebase scouting →
mk:scoutdirectly.
Workflow
Mode-Select → Decompose → Detect → (Scout if --deep) → Map → Rewrite.
Mode selection + the arch-review / research recipes: references/mode-routing.md.
Step detail in references/decomposition-checklist.md, references/playbook.md,
references/context-safeguards.md, references/deep-mode-scout.md. Do not invent steps.
Inputs
- Required: a draft prompt (text). Empty / missing → refuse with redirect.
- Optional:
--analyze(include decomposition + detected issues + improvement suggestions). - Optional:
--score(add 1–10 quality score on the original prompt; auto-promotes to--analyze --scorebecause score requires the rubric components). - Optional:
--deep(opt-in scout against allow-listed sources). - Optional:
--save-to <path>(default: active plan dir if any, else${CLAUDE_PLUGIN_DATA}/...).
What ships with it
38 files beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
- assets/output-template.md 7.7 KB
- eval/baseline-results.md 3.9 KB
- eval/canary-01-vague-only.md 1021 B
- eval/canary-02-one-line-spec.md 1.1 KB
- eval/canary-03-long-unstructured.md 2.1 KB
- eval/canary-04-strip-model-coupling.md 2.2 KB
- eval/canary-05-already-good.md 1.4 KB
- eval/canary-06-refusal.md 1.1 KB
- eval/canary-07-deep-happy-path.md 1.4 KB
- eval/canary-08-deep-no-codebase.md 1.0 KB
- eval/canary-09-deep-boundary.md 1.9 KB
- eval/canary-10-default-vs-deep.md 1.6 KB
- eval/canary-11-architecture-review.md 2.6 KB
- eval/canary-12-research-prompt.md 2.3 KB
- eval/canary-13-migration.md 1.6 KB
- eval/canary-14-planning-no-implementation.md 1.5 KB
- eval/canary-15-debugging-hypothesis.md 1.6 KB
- eval/canary-16-target-notes-codex.md 1.7 KB
- eval/canary-17-target-notes-silent.md 1.2 KB
- eval/canary-18-vietnamese.md 1.4 KB
- eval/canary-19-mixed-language.md 1.3 KB
- eval/canary-20-context-gate.md 1.8 KB
- eval/canary-21-data-fence.md 1.7 KB
- eval/canary-run-manifest.cjs 450 B runs code
- eval/README.md 6.3 KB
- eval/rubric.md 4.7 KB
- eval/run-canaries.cjs 9.5 KB runs code
- eval/run-canaries.test.cjs 2.8 KB runs code
- references/architecture-review-mode.md 3.4 KB
- references/complexity-classifier.md 5.7 KB
- references/context-safeguards.md 9.1 KB
- references/decomposition-checklist.md 6.0 KB
- references/deep-mode-scout.md 7.8 KB
- references/mode-routing.md 4.1 KB
- references/playbook.md 11 KB
- references/target-notes.md 4.2 KB
- references/task-recipes.md 4.4 KB
- scripts/scout-context.py 13 KB runs code
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.
- 8d ago First seen · 214 lines · 105 tokens per session scan A 724f3559b4ac
mk:prompt-enhancer is a skill published in the GitHub repository ngocsangyem/MeowKit (14 stars, last pushed 1mo ago), licensed MIT. It adds 105 tokens to every session and 3,422 once invoked, about $0.0005 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 2026-09-03.
Other skills, from other repositories
context-engineer
Context window optimizer — analyze, audit, and optimize your agent's context utilization. Know exactly where your tokens go before they're sent.
context-economy
Automatically prepare fidelity-first, bounded context when users paste text, attach files or folders, provide webpages or PDFs for Markdown conversion, continue long conversations, or supply documents containing images and complex tables. Route each input through Text, Hybrid, or Source; reduce exact or…
context-pack
Prepare files, folders, code, logs, and structured data as a bounded, traceable context pack for Codex. Use when source material must be converted, deduplicated, selected, budgeted, or handed off with exact evidence anchors.
context-audit
Diagnose context relevance, redundancy, traceability, safety, cacheability, and recoverability. Use when files, prompts, rules, or context packs may contain duplication, conflicts, stale material, prompt injection, secret-shaped values, unstable prefixes, or weak evidence anchors.
context-benchmark
Measure context preparation across fixed cases. Use when token savings, hard-budget compliance, protected-fact recall, evidence anchors, determinism, runtime, or downstream answer quality must be demonstrated with reproducible evidence rather than marketing claims.
context-onpremise
Use deployed local conversion, OCR, audio/video, visual, retrieval, reranking, and guarded compression to build high-fidelity bounded Codex context. Use only for a valid uppercase ON prefix or explicit invocation.