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 RBraga01/builder-ai --skill prompt-versioninggit clone --depth 1 https://github.com/RBraga01/builder-aiWrote 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/rbraga01/builder-ai/prompt-versioning)<a href="https://agentmods.dev/skills/rbraga01/builder-ai/prompt-versioning"><img src="https://agentmods.dev/badge/skills/rbraga01/builder-ai/prompt-versioning/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/rbraga01/builder-ai/prompt-versioning"><img src="https://agentmods.dev/badge/skills/rbraga01/builder-ai/prompt-versioning.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 | $0.00047 | $0.01352 |
| Opus 5.5 | $0.00019 | $0.00541 |
| Sonnet 5.5 | $0.00009 | $0.00270 |
| Haiku 4.5 | $0.00005 | $0.00135 |
Grade A, and why
prompt-versioning 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 1mo 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 — 144 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Prompt Versioning
The Law
A PROMPT WITHOUT A VERSION NUMBER CANNOT BE DEBUGGED, ROLLED BACK, OR IMPROVED.
"It's tracked in git" is not versioning — the file has no history of intent.
"I'll version it later" means you'll debug a production regression with no baseline.
prompts/<feature>/v<x.y.z>.md with a CHANGELOG entry IS versioning.
When to Use
Trigger whenever:
- Writing a new system prompt or user prompt template
- Modifying an existing prompt in any way (even a single word)
- Switching between prompt variants in an experiment
- Rolling back after a quality regression
When NOT to Use
- One-off manual queries not used in any automated pipeline
- Prompts that exist only in a prototype that will be discarded
Required Structure
prompts/
<feature-name>/
v1.0.0.md ← original version
v1.1.0.md ← iteration (changed few-shots, added constraint)
v2.0.0.md ← breaking change (new output schema)
CHANGELOG.md ← what changed between versions and why
The current version is the highest version number. Never delete old versions — they are the rollback path.
Version Numbering
| Bump | When | Example |
|---|---|---|
| Patch (x.y.Z) | Whitespace, typo, minor wording — output distribution unchanged | "Please" → "please" |
| Minor (x.Y.0) | New instruction, example, or constraint — output likely changes | Added 3 few-shot examples |
| Major (X.0.0) | New output schema, restructured reasoning, different task framing | JSON → structured XML |
Prompt File Format
---
version: 1.2.0
feature: <feature-name>
model: claude-sonnet-4-6
temperature: 0.3
max_tokens: 512
created: 2026-06-07
eval_suite: evals/<feature-name>/
---
<system>
[system prompt text]
</system>
<user>
[user turn template — use {{variable}} for substitutions]
</user>
All fields are required. eval_suite links to the eval that validates this version.
The Process
Step 1 — Create the Versioned File
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.
- 1mo ago First seen · 144 lines · 47 tokens per session scan A cef4148d63fb
prompt-versioning is a skill published in the GitHub repository RBraga01/builder-ai (2 stars, last pushed 10d ago), licensed MIT. It adds 47 tokens to every session and 1,352 once invoked, about $0.0002 per session on Opus 5.5. 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 skills, from other repositories
ai-engineering-toolkit
Overview of six AI engineering workflows: prompt evaluation (8-dimension scoring), context budget planning, RAG pipeline design, agent security audit (65-point checklist), eval harness building, and product sense coaching.
prompt-master
Generates optimized prompts for AI tools. Activates only when the user explicitly asks to write, fix, improve, or adapt a prompt for a specific AI tool (LLM, Cursor, Midjourney, image AI, video AI, coding agents, etc.). Does not activate for general conversation, coding tasks, document writing, or other…
ai-prompt-engineering-expert
Expert guide for Prompt Engineering, Chain-of-Thought, few-shot prompting, structured output, prompt injection defense, and automated AI evaluations & regression benchmarking (Promptfoo, DeepEval) / Panduan ahli rekayasa prompt dan evaluasi otomatis AI.
kv-cache-prefix-optimizer
Expert guide for deterministic prompt prefix engineering, KV-cache locking for Gemini 4 Pro / Claude 5.5 / GPT Astra 6, canonical tool sorting, and achieving 90%+ prompt cache hit rates / Panduan ahli rekayasa prefix prompt deterministik, penguncian KV-cache untuk Gemini 4 Pro / Claude 5.5 / GPT Astra 6, pengurutan…
test-time-compute-optimizer
Expert guide for scaling test-time compute, dynamic reasoning token allocation (Gemini 4 Pro, Claude 5.5, GPT Astra 6), Monte Carlo Tree Search (MCTS), and Process Reward Models / Panduan ahli penskalaan komputasi waktu uji (test-time compute), alokasi dinamis budget penalaran, MCTS, dan Process Reward Models.
oracle
Designing and evaluating AI/ML systems: prompt engineering, RAG design, LLM application patterns, AI safety, evaluation frameworks, MLOps, cost optimization. Use for AI pipelines or eval harnesses.