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 gul-labs/craftsman-marketplace --skill craft-aigit clone --depth 1 https://github.com/gul-labs/craftsman-marketplaceWrote 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/gul-labs/craftsman-marketplace/craft-ai)<a href="https://agentmods.dev/skills/gul-labs/craftsman-marketplace/craft-ai"><img src="https://agentmods.dev/badge/skills/gul-labs/craftsman-marketplace/craft-ai/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/gul-labs/craftsman-marketplace/craft-ai"><img src="https://agentmods.dev/badge/skills/gul-labs/craftsman-marketplace/craft-ai.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 | $0.00129 | $0.02971 |
| Opus 5 | $0.00064 | $0.01486 |
| Sonnet 5 | $0.00026 | $0.00594 |
| Haiku 4.5 | $0.00013 | $0.00297 |
Grade A, and why
craft-ai 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 — 190 lines — stays where its author put it; the contents beside it link to each section on GitHub.
AI Craft
This skill encodes one engineer's standard for shipping LLM-powered features safely, applied the same way across every repo. The method and opinions live here; the project specifics (which provider, which SDK, which framework) live in the target repo's code and config — always discover them, never assume or hardcode.
Operating principle — discover before you build
Different repos already have different pieces in place. Before changing anything, spend a few minutes mapping the current posture so you extend rather than conflict:
package.json/ lockfile / requirements → which LLM SDK is present (OpenAI, Anthropic, Vercel AI SDK, LangChain, LlamaIndex)? Which is the default model tier?grepfor the API key — is it read through a validated server-side env schema, or does it leak into aNEXT_PUBLIC_*/ client-bundled / mobile-app constant?- Find every call site that sends a prompt — what user-controlled or retrieved content reaches the system prompt or context window, and is there any structural separation between instructions and untrusted content?
- Check whether the integration uses tool-use / function-calling — what can the model actually invoke, and is there a confirmation step before a consequential action fires?
- Look for logging/observability around LLM calls — are raw prompts and completions logged, and where do those logs live?
- Check for rate limits,
max_tokenscaps, and timeouts on LLM-calling routes; check for retry logic around tool actions. - Look for any eval harness, golden test set, or regression check tied to prompt changes.
State what you found, then propose the smallest set of changes that closes the gaps.
The AI layers (work in this order)
- Prompt-injection surface — anything that reaches the model's context that isn't the
developer's own instructions (user input, retrieved documents, scraped pages, tool output) is
a potential injection vector. Structural separation between instructions and untrusted content,
allow-listed tools, and human confirmation for consequential actions are the mitigations — none
of them fully close the gap. See
references/prompt-injection.md. - Keys & spend — provider keys never ship to a client bundle or mobile app; every LLM-calling
route has a rate limit, a
max_tokenscap, and a bound on loop/agent iterations; streaming requests abort when the client disconnects. Seereferences/keys-and-spend.md. - Data privacy — inventory what PII leaves the building in a prompt to a third-party API,
check the provider's retention/training posture, and scrub prompts/completions before they hit
observability logs. See
references/data-privacy.md. - Reliability & evals — provider outages are routine, so every call has a timeout and a
fallback; non-idempotent tool actions are never blindly retried; a minimal golden-case eval
harness catches prompt regressions before they ship. See
references/reliability-evals.md.
What ships with it
4 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.
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 Changed · +17 lines 45eb975ddf57
- 12d ago First seen · 173 lines · 129 tokens per session scan A 133dfadf58f1
craft-ai is a skill published in the GitHub repository gul-labs/craftsman-marketplace (1 stars, last pushed today), licensed MIT. It adds 129 tokens to every session and 2,971 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 2026-08-31.
Other skills, from other repositories
ai-workflow-architect
Designs AI systems, automations, and agent workflows for a business — identifying which manual work is worth automating, how to structure the system, which tools fit, and what could go wrong. Use this to automate part of an operation, design an agent or MCP workflow, reduce repetitive manual work, connect tools into a…
prompt-tuning
Tune a prompt, or anything whose quality is measured by non-deterministic model output, without chasing noise - a noise baseline before the first edit, medians over repeated runs, enforcement AFTER generation rather than in the wording. Use when iterating on prompts or model-judged output.
laravel-ai-sdk
Use when integrating AI agents, tool calling, embeddings, structured output, or streaming in Laravel 13 via the laravel/ai package.
prompt-library
Use when looking for a ready-made agent or task template (code review, support, data analysis, translation, extraction) instead of writing one from scratch.
prompt-testing
Use when comparing two prompt variants, defining quality/efficiency/robustness metrics, or deciding whether to adopt a challenger prompt over a baseline.
prompt-creation
Use when writing a new system prompt from scratch, structuring it with Anthropic's 9-element format, or choosing emphasis levels and few-shot examples.