Need a Canonical Agent-Ready Developer Platform Pattern
Developer platforms need a canonical pattern for what 'agent-ready' actually means — a spectrum from markdown and llms.txt at the simple end through full discover-sign-up-and-build-PoC agent journeys at the complex end.
Take Control Of Your Signals — Become a Naftiko Design Partner Today!
Persona Story:
Maya is trying to make her developer platform agent-ready and there is no single pattern to point at. The simple end is clear — markdown docs, llms.txt, handoff instructions to AI — but the complex end (agent discovers the platform, signs up, builds proofs of concept across providers, returns to a human with options) is undefined. Test harnesses and evals are being built one platform at a time. The pattern needs to be canonical, not bespoke.
Problem Context
- DevEx leads are publishing posts framing agent-readiness as a spectrum from simple (markdown / llms.txt) to complex (full agent journey)
- The complex end is still aspirational — most platforms don’t have an agent that can discover, sign up, and build a PoC unattended
- Each platform team is building test harnesses and evals from scratch to measure whether prompts and agent journeys actually work
- A canonical pattern would shorten the path for every developer platform team simultaneously
Problem Impact
- Each platform team independently invents the same agent-readiness primitives — markdown structure, MCP surface, skills exposure, test harness
- Developers cannot compare platforms on agent-readiness because there is no shared definition
- Agent reliability across platforms stays uneven because every team measures it differently
- The agent-readiness conversation stalls at the simple end because the complex end has no reference implementation
Naftiko Today
- Executable YAML capability specs are a canonical pattern for the structured-content side of agent-readiness — every capability is the same shape, parsed deterministically
- MCP exposure (Streamable HTTP, stdio) gives developer platforms a single canonical surface for agent access
- Agent Skills exposure publishes capabilities as structured skills that fit the complex end of the spectrum (discover, plan, execute)
- JSON Schema validation and Spectral ruleset (15 rules) give platforms a shared rubric for what an agent-ready capability looks like
Naftiko Tomorrow
- HTML / Markdown format support (Second Alpha) will cover the simple end of the spectrum (markdown / llms.txt) inside the same capability framework as the complex end
- Tool annotations (Second Alpha) will let platforms declare behavioral metadata (readOnly / destructive / idempotent) that agent journeys need at the complex end
- Starter templates (Second Alpha) will publish reference implementations of the agent-ready developer platform pattern so teams stop reinventing
- Naftiko Shipyard MVP (Fleet Second Alpha) will give DevEx teams the workbench where agent-readiness is assembled and measured consistently