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 skills/aegntic/compound-engineering/agent-native-architecturenpx skills add aegntic/compound-engineering --skill agent-native-architecturegit clone --depth 1 https://github.com/aegntic/compound-engineeringWrote 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/aegntic/compound-engineering/agent-native-architecture)<a href="https://agentmods.dev/skills/aegntic/compound-engineering/agent-native-architecture"><img src="https://agentmods.dev/badge/skills/aegntic/compound-engineering/agent-native-architecture.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 | $0.00050 | $0.01018 |
| Opus 5 | $0.00025 | $0.00509 |
| Sonnet 5 | $0.00010 | $0.00204 |
| Haiku 4.5 | $0.00005 | $0.00102 |
Grade A, and why
agent-native-architecture 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 4d 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 — 93 lines — stays where its author put it; the contents beside it link to each section on GitHub.
agent-native-architecture Skill
Design software where agents are first-class operators, not an afterthought. Start with parity, primitive tools, and explicit contracts so new features can be delivered as prompts, not hard-coded workflows.
When to use
- Designing or refactoring an agent-native application.
- Creating MCP tools, tool contracts, or system prompts.
- Reviewing action parity between UI and agent capabilities.
- Producing the architecture artifact consumed by the compound workflow chain.
Workflow
- Identify the user's goal: design, files/workspace, tool design, domain tools, execution patterns, system prompts, context injection, action parity, self-modification, product design, mobile patterns, testing, or refactoring.
- Load only the reference files needed for that topic, then map the advice onto the current product, data model, and deployment constraints.
- Run the architecture checklist below before recommending new abstractions.
- If this is part of the compound workflow chain, turn the guidance into a concrete architecture artifact under
docs/architecture/.
Core principles
- Parity: if the user can do it, the agent can achieve the same outcome.
- Granularity: tools are primitives; features are prompt-defined outcomes.
- Composability: new capabilities should come from new prompts before new code.
- Emergent capability: open-ended requests should reveal gaps and product demand.
- Improvement over time: context, prompt refinement, and safe self-modification should make the system better with use.
Reference router
- Design from scratch ->
references/architecture-patterns.md - Files and workspace design ->
references/files-universal-interface.md,references/shared-workspace-architecture.md - MCP and primitive tool design ->
references/mcp-tool-design.md - Domain-tool thresholds ->
references/from-primitives-to-domain-tools.md - Execution and completion loops ->
references/agent-execution-patterns.md - System prompt design ->
references/system-prompt-design.md - Dynamic context injection ->
references/dynamic-context-injection.md - Action parity reviews ->
references/action-parity-discipline.md - Self-modification ->
references/self-modification.md - Product implications ->
references/product-implications.md - Mobile patterns ->
references/mobile-patterns.md - Testing ->
references/agent-native-testing.md - Refactoring legacy systems ->
references/refactoring-to-prompt-native.md
What ships with it
14 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/action-parity-discipline.md 11 KB
- references/agent-execution-patterns.md 13 KB
- references/agent-native-testing.md 16 KB
- references/architecture-patterns.md 17 KB
- references/dynamic-context-injection.md 9.4 KB
- references/files-universal-interface.md 9.9 KB
- references/from-primitives-to-domain-tools.md 12 KB
- references/mcp-tool-design.md 15 KB
- references/mobile-patterns.md 25 KB
- references/product-implications.md 13 KB
- references/refactoring-to-prompt-native.md 8.4 KB
- references/self-modification.md 7.7 KB
- references/shared-workspace-architecture.md 20 KB
- references/system-prompt-design.md 6.4 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.
- 4d ago First seen · 93 lines · 50 tokens per session scan A 3b48aea94bd0
agent-native-architecture is a skill published in the GitHub repository aegntic/compound-engineering (2 stars, last pushed 2mo ago), licensed MIT. It adds 50 tokens to every session and 1,018 once invoked, about $0.0003 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
methodology
Analyzes captured HTTP traffic, designs the CLI architecture, and implements the Python CLI package (Phase 2): parse raw-traffic.json, identify the protocol, write api-spec.json, scaffold from templates, and implement endpoint methods and Click command groups. Use after a capture completes and raw-traffic.json exists.
standards
Runs Phase 4 review/publish/verify for a cli-web- CLI: implementation review by 3 parallel agents, the tiered quality checklist (Tier 1 critical fail-fast, then comprehensive), pip install + smoke test, and per-CLI skill generation. Use when a CLI's tests pass and it is ready to be validated and published.
testing
Writes and documents the test suite for a generated cli-web- CLI (Phase 3): unit tests with mocked HTTP, live E2E tests, subprocess tests via resolvecli, and the TEST.md plan/results record. Use after the methodology skill completes implementation.
sync-check
Walks the documentation dependency web after a CLI code change and reports which downstream files (skills, SOPs, READMEs, plugin references) are out of sync. Use after fixing a bug, adding a command, changing auth behavior, refactoring, or before committing — and when the user says "sync check", "update docs"…
boilerplate
Documents the template inventory and variable contract behind scaffold-cli.py — which Jinja2 template renders with which variables for each site profile. Use during Phase 2 scaffolding when choosing scaffold flags or understanding what the generated boilerplate contains. The scaffold-cli.py script is the primary path.
gap-analyzer
Compares a CLI's implemented commands against its APP.md API map and traffic-analysis.json to find missing endpoints, incomplete CRUD, dead client methods, and priority gaps. Runs as the mandatory first step of /cli-anything-web:refine and as an optional pre-review scan in standards.