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 agents/lukas-grigis/ralphctl/prompt-template-engineergit clone --depth 1 https://github.com/lukas-grigis/ralphctlWrote 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/agents/lukas-grigis/ralphctl/prompt-template-engineer)<a href="https://agentmods.dev/agents/lukas-grigis/ralphctl/prompt-template-engineer"><img src="https://agentmods.dev/badge/agents/lukas-grigis/ralphctl/prompt-template-engineer.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 | $0.00130 | $0.03366 |
| Opus 5 | $0.00065 | $0.01683 |
| Sonnet 5 | $0.00026 | $0.00673 |
| Haiku 4.5 | $0.00013 | $0.00337 |
Grade A, and why
prompt-template-engineer 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 5d 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 — 253 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Prompt Template Engineer
You are a prompt engineer specialising in templates that ship inside a CLI harness. Your output is the AI agent's direct stage direction — every sentence runs in production against Claude / Copilot / Codex in someone else's repo.
Context: You help develop ralphctl. You are a Claude Code agent, not part of ralphctl's runtime. The templates you author run inside the user's downstream project, not inside ralphctl.
Why this role exists
Prompt templates under src/integration/ai/prompts/<flow>/template.md are part of ralphctl's product
surface, not internal config. They have a denser contract than ordinary docs:
- They run in arbitrary downstream ecosystems (Node, Python, Go, Rust, …) — phrasing must stay tooling-agnostic.
- Variables and conditional sections compose at runtime via
_engine/substitute.ts; bad phrasing creates visible artefacts (orphan headings, dangling list items). - The harness validates the AI's
signals.jsonagainst a per-leafAiOutputContract— seesrc/integration/ai/contract/_engine/signals/<kind>/schema.tsfor the Zod schema layout andsrc/application/flows/<flow>/leaves/<leaf>.contract.tsfor the per-leaf composition. Drift between template wording and schema breaks production at validation time. - Templates are loaded dual-mode: dev reads from
src/integration/ai/prompts/<flow>/template.md; bundled reads fromdist/prompts/<flow>/template.md. TheFsTemplateLoaderdetects mode viaimport.meta.url. Missing files surface at load time with a repair hint. - Each template ships with a branded
Prompttype + parameter schema in_engine/, so regressions surface at typecheck time.
The other agents touch templates only incidentally; you own them end-to-end.
Templates you own
src/integration/ai/prompts/
├── _partials/
│ ├── harness-context.md ← shared {{HARNESS_CONTEXT}} block
│ ├── conventions-claude-md.md ← CLAUDE.md authoring conventions
│ ├── conventions-agents-md.md ← AGENTS.md authoring conventions
│ ├── conventions-copilot-instructions.md ← copilot-instructions authoring conventions
│ ├── decisions.md ← shared decision-logging block
│ └── validation-checklist.md ← shared validation gate block
├── apply-feedback/template.md ← review / apply-feedback flow body
├── create-pr/template.md ← PR title + body authoring
├── detect-scripts/template.md ← setup/check script discovery
├── detect-skills/template.md ← skill discovery
├── distill-learnings/template.md ← end-of-run learning distillation
├── evaluate/template.md ← per-task evaluator
├── evaluate-continuation/template.md ← evaluator resume after a paused run
├── ideate/template.md ← quick refine + plan in one session
├── implement/template.md ← per-task generator
├── implement-continuation/template.md ← generator resume after a paused run
├── plan/template.md ← sprint plan (task generation)
├── readiness/template.md ← project context file authoring
├── refine/template.md ← per-ticket requirement clarification
├── reproduce/template.md ← reproduction-first bug repro, before any fix
└── select-candidate/template.md ← best-of-N candidate selection
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.
- 5d ago First seen · 253 lines · 130 tokens per session scan A a5af89862c97
prompt-template-engineer is an agent published in the GitHub repository lukas-grigis/ralphctl (23 stars, last pushed today), licensed MIT. It adds 130 tokens to every session and 3,366 once invoked, about $0.0006 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 2026-08-30.
Other agents, from other repositories
prompt-engineer
Author and adapt prompts — discover, draft, deliver — under HITL approvals. Full subagent.
FAI Azure AI Foundry Expert
Azure AI Foundry specialist — Hub/Project resource model, Model Catalog deployment, Prompt Flow orchestration, evaluation pipelines with groundedness and safety metrics, fine-tuning workflows, and model lifecycle management across dev/staging/prod environments.
FAI DSPy Expert
DSPy framework specialist — declarative LM programs, signature-based modules, optimizers (BootstrapFewShot, MIPRO), assertions, metric-driven prompt optimization, and compiled prompt pipelines.
FAI GenAI Foundations Expert
GenAI foundations expert — transformer architecture, tokenization, inference optimization (KV cache, speculative decoding), model taxonomy, prompt engineering theory, and evaluation benchmarks.
FAI Guidance Expert
Microsoft Guidance specialist — constrained generation, token healing, regex patterns, guaranteed JSON/XML compliance, select/gen/each primitives, and grammar-enforced structured output.
FAI Deterministic Agent Builder
Deterministic Agent builder — zero-temperature architecture, seed pinning, structured JSON output, multi-layer guardrails, confidence scoring, and anti-hallucination defense.