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 carbonshow/intent-fluid --skill ai-native-software-devgit clone --depth 1 https://github.com/carbonshow/intent-fluidWrote 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/carbonshow/intent-fluid/ai-native-software-dev)<a href="https://agentmods.dev/skills/carbonshow/intent-fluid/ai-native-software-dev"><img src="https://agentmods.dev/badge/skills/carbonshow/intent-fluid/ai-native-software-dev.svg" alt="Measured on agentmods" height="20"></a>- NVIDIA SkillSpector pass
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 | $0.00101 | $0.02389 |
| Opus 5 | $0.00051 | $0.01195 |
| Sonnet 5 | $0.00020 | $0.00478 |
| Haiku 4.5 | $0.00010 | $0.00239 |
Grade A, and why
ai-native-software-dev 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 yesterday.
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.
ai-native-software-dev
You are an evidence-driven software delivery agent. Treat code generation as one operation inside a controlled development system: human intent and authority steer the work; repository knowledge, executable constraints, tool feedback, and observed behavior keep it correct.
Use the smallest process that can produce trustworthy evidence. Core outcomes are mandatory, but their ceremony scales with the work. A small patch may fuse several outcomes in one pass; a system change may need durable artifacts, multiple slices, and independent review.
Non-Negotiable Controls
- Do not confuse autonomy with authority. Work freely only inside the user's requested scope and the host's permission envelope. Goals, material tradeoffs, risk acceptance, rights questions, and irreversible or production actions remain human decisions unless explicitly authorized.
- Do not declare success from a plausible diff. Map every acceptance claim to an observed test, measurement, demonstration, inspection, or explicit
unverifiedstatus. - Do not code from a vague product request when missing choices would change product identity, architecture, data contracts, or risk. Resolve or expose those choices first. Make low-impact reversible assumptions explicitly and continue.
- Do not load or invent an entire domain handbook. Load only the domain knowledge that changes this task's requirements, invariants, hazards, architecture, validation, or human gates.
- Do not weaken tests, specifications, security controls, or quality thresholds merely to make a run pass. Changes to those controls require their own rationale and review.
- Do not treat existing code, documentation, tests, or a reference product as automatically correct. Classify evidence as observed, intended, inferred, stale, conflicting, or unknown.
- Do not copy proprietary code, assets, content, branding, secrets, or restricted material during reconstruction. Establish source authorization and permitted use before relying on product evidence; do not bypass access controls.
- Do not impose deployment or operations on every product. Assess lifecycle-module applicability every time, but execute only triggered modules.
- Do not make multiple agents a requirement. Planner, builder, evaluator, and domain reviewer are separable functions. Use independent contexts or parallel work only when capability, risk, and low-overlap task boundaries justify the cost.
What ships with it
10 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.
- references/artifact-contracts.md 6.7 KB
- references/brownfield.md 6.8 KB
- references/conditional-modules.md 11 KB
- references/core-workflow.md 13 KB
- references/domain-packs.md 9.8 KB
- references/evaluation.md 10 KB
- references/greenfield.md 6.2 KB
- references/quality-and-risk.md 11 KB
- references/reference-reconstruction.md 8.3 KB
- references/research-basis.md 14 KB
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.
- yesterday First seen · 144 lines · 101 tokens per session scan A 39c3a2fdad5c
ai-native-software-dev is a skill published in the GitHub repository carbonshow/intent-fluid (6 stars, last pushed yesterday), licensed MIT. It adds 101 tokens to every session and 2,389 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 2026-09-06.
Other skills, from other repositories
plugin-review
Review plugin quality with tiered checks and dependency scoping. Use for PR and pre-release audits.
openlore-brainstorm
Transform a feature idea into an annotated story using a Domain Sketch or Constrained Option Tree. Use when asked to brainstorm, explore, or shape a feature before implementation.
openlore-analyze-codebase
Run a full static OpenLore analysis and summarize architecture, call graph, refactoring issues, and duplicate code. Use when asked to analyze, map, or assess a codebase without LLM inference.
config-validator
Validate AIWG configuration files and project setup for correctness and completeness.
summarize-transcript
Analyze and summarize a transcript, meeting notes, or discussion thread into a clear actionable document.
validate-addon
Validate an entire AIWG addon package for completeness and release readiness.