~/agent-sdlcAgent SDLC · Early ecosystemGitHub ↗

~/agentsdlc

Agent SDLC

Open artefacts for
agentic software engineering.

Agents move between sessions, tools and repositories. Engineering intent and reasoning should survive them.

See how Agent SDLC works ↓

STANDARDS

STANDARD

ORBIT

Keep execution aligned
with durable outcomes.

Planning and Reconciliation Specification

Define intent and outcomes once. Reconcile them against current software reality. Materialise only the next bounded execution slice.

Explore ORBIT →

orbitspec.dev

STANDARD

AgDR

Record what the agent
decided and why.

Agent Decision Record Specification

Preserve consequential technical decisions and rationale beyond the agent session that created the code.

Explore AgDR →

agdr.dev

HOW THEY WORK TOGETHER

Independent by design.
Composable when useful.

INTEROPERABILITY

ORBIT determines what work is justified next. AgDR preserves why a consequential technical decision was made during that work.

Neither standard depends on the other.

ORBIT Plan
    ↓
Reconciliation
    ↓
Execution Slice
    ↓
Agent
 ├──→ Code
 └──→ AgDR
{
  "rel": "implements",
  "type": "orbit.execution-slice",
  "id": "slice-004"
}

WHY AGENT SDLC

Sessions disappear.
Engineering context should not.

Agentic development creates useful context inside conversations, agent runs and prompts. Agent SDLC defines small durable artefacts for the parts that need to survive.

TransientDurable artefacts

Conversation
Useful while the session is active, but not a durable engineering record.

ORBIT Plan
Preserves intended outcomes.

Agent session
Ends when the run finishes or context resets.

AgDR
Preserves why a consequential technical decision was made.

Runtime prompt
Describes what the agent knew or was told at that moment.

Evidence and execution provenance
Can preserve what supported an assessment and which project state justified work.

OPEN BY DEFAULTVersioned specificationsMachine-readable schemasPublic change historyImplementation-neutral contractsAbout Agent SDLC →