agentic-build-loop

A step-by-step method for directing coding agents to build software from a written specification and acceptance tests. It covers planning, implementation, review, and checking the result.

In plain words
What is it for?
Use it to define requirements, plan the code structure, guide agent iterations, and verify features with tests.
Why use it?
It reduces the risk of an agent building the wrong feature, producing tangled code, or stopping at code that merely appears to work.

Skill for Claude CodeCodex

Part of the ai-native-os plugin — 25 skills, 6 agents shipped together

Install

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.

agentmods
npx agentmods add skills/impactbrussels/ainativeos/agentic-build-loop
Any agent
npx skills add impactbrussels/AINativeOS --skill agentic-build-loop
Clone the repo
git clone --depth 1 https://github.com/impactbrussels/AINativeOS

Made for: Claude Code, Codex.

Or install ai-native-os, the plugin that ships this one along with the rest of its 25 skills, 6 agents.

Per session 174 Skills are progressive disclosure: only the name and description are preloaded; the body loads when the skill is used.
When invoked 1,130 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. Scan, not verified.
Origin original No closer match found in the catalogue.
Token cost

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.

ModelPer sessionOnce invoked
Fable 5 $0.00174 $0.01130
Opus 5 $0.00087 $0.00565
Sonnet 5 $0.00035 $0.00226
Haiku 4.5 $0.00017 $0.00113

Measured 2d ago against content hash 0568b8b0bf93, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

agentic-build-loop 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 2d 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.

skills/agentic-build-loop/SKILL.md · 82 lines

How it starts

The opening of the file, as written. The whole thing — 82 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Agentic Build Loop

The cost of syntax is zero. An agent writes correct code faster than you can, in any language, so the job is not to type lines. You specify the system and the tests that judge it, and the agents iterate until the suite goes green. Hand them a vague request and they build the wrong thing fluently. Hand them a spec and a harness and they have nowhere to hide. You are the conductor; the architecture is the score.

The method

Full build loop, eval gates, a worked build, anti-patterns and a copyable template: references/agentic-build-method.md.

Step 1: Write the spec and the acceptance tests first

Before any code, write two things the agent cannot infer: what the feature must do, and the handful of tests that decide whether it did. Plain language is fine if that is all you have. The tests are the contract the agent builds to. No target, no grip, and the agent drifts while you call the drift a feature.

Step 2: Plan mode, framework first

Run the agent in plan mode and ask for the structure, not the logic: the modules, their boundaries, their interfaces, the skeleton empty of muscle. Read that hierarchy against your architecture and validate it before a single line of logic. You can move a wall on a diagram for nothing. You cannot move it once the roof is on.

Step 3: Let agents fill the logic

Only when the frame is right, release the agents to write the logic. They run the tests, read the failures, fix, and run again until the suite passes. Review every diff. A passing demo proves the agent typed something plausible, not that it built the right thing.

Step 4: Keep the master file current

The agents read one master file every session (CLAUDE.md, AGENTS.md, .cursor/rules, GEMINI.md). At the end of the session, append the trade-offs you made: the boundary you moved, the constraint you added, the thing an agent must never do next time. The file is where the multiplier lives.

Output

Read the full file on GitHub · 82 lines

Files

What ships with it

1 file 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.

Changes

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.

  1. 2d ago First seen · 82 lines · 174 tokens per session scan A 0568b8b0bf93

Subscribe to this mod's changes

agentic-build-loop is a skill published in the GitHub repository impactbrussels/AINativeOS (1 stars, last pushed 2mo ago), licensed Apache-2.0. It adds 174 tokens to every session and 1,130 once invoked, about $0.0009 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.

Related

Other skills, from other repositories

external-context

Invoke parallel document-specialist agents for external web searches and documentation lookup.

Yeachan-Heo/oh-my-claudecode · 16 tokens

security-ownership-map

Analyze git repositories to build a security ownership topology (people-to-file), compute bus factor and sensitive-code ownership, and export CSV/JSON for graph databases and visualization. Trigger only when the user explicitly wants a security-oriented ownership or bus-factor analysis grounded in git history (for…

HKUDS/DeepCode · 99 tokens

moai-ref-ui-polish

UI polish and interface-completion reference: the small visual details — concentric border radius, optical alignment, shadow-vs-border, motion easing, typography smoothing, tabular numbers, icon stroke weight, hit areas — that separate polished interfaces from generic ones. Agent-extending skill that amplifies…

modu-ai/moai-adk · 98 tokens

submit-github-pr

Use when publishing an existing TensorRT-Model-Connect change as a GitHub pull request. Verifies authenticated repository access, branch and diff scope, validation evidence, commit identity, reviewer-facing text, exact pushed head, and the created draft PR without merging it.

NVIDIA/TensorRT-Model-Connect · 58 tokens

lemlist-campaign-from-icp

Use when the user says "create a lemlist campaign for X", "build a lemlist campaign from this ICP", "spin a lemlist campaign for Y", "ICP to lemlist", "natural language to lemlist campaign", "draft a lemlist campaign", "lemlist campaign for VPs/Managers/ICs at Z", or any variant indicating they want to turn a…

Othmane-Khadri/YALC-the-GTM-operating-system · 251 tokens

cross-task-learner

Enable agent loops to learn from similar past tasks and share patterns across loops.

jmagly/aiwg · 20 tokens