Work with your coding agent. Keep the project in control.
Say what you want in normal language. Skopos gives the conversation reliable project context, a real work boundary, approved actions, proof, and a path into the next Session.
Start with what you want to accomplish.
Each path shows a natural prompt and the project responsibility Skopos keeps stable behind it.
Ask what is happening in the project
Return after time away without replaying old conversations or loading the whole repository.
“Bring me up to speed. Show active work, material decisions, recent proof, and the next safe action.”
Skopos selects the current Task and the Project Memory relevant to it.
Plan and build one feature
Carry one outcome from the first discussion through bounded implementation, proof, and Memory review.
“Help me build this feature. Recover relevant project truth, clarify meaningful decisions, and bound the work before editing.”
Skopos keeps intent, ownership, checks, Evidence, and closure connected.
Record a decision that should last
Keep a product or architecture choice available after the conversation ends.
“Record this accepted decision in its canonical project owner. Keep the reasoning and consequences, and do not duplicate existing truth.”
Skopos distinguishes durable Project Memory from temporary Task state and agent inference.
Start work with a real boundary
Turn an agreed outcome into observable acceptance, owned paths, constraints, and non-goals.
“Create one bounded Task. Show me its acceptance criteria, narrowest owned paths, constraints, and non-goals before implementation.”
Skopos protects the Task boundary and reports work that falls outside it.
Continue in a fresh Session
Move current work into a clean coding-agent conversation without pretending the handoff was delivered.
“Prepare this active Task for a fresh Session. Preserve current intent and live project state, then show me the exact handoff before delivery.”
Skopos verifies, renders, delivers or exposes a manual fallback, then binds the receiving Session.
Split genuinely independent work
Give separate agents explicit child outcomes while the original conversation remains the reviewer.
“Split only independent parts. Give each child acceptance and owned paths, and keep this conversation responsible for combined review.”
Skopos coordinates child authority, path claims, dependencies, and parent closure.
Prove the work before calling it done
Select the checks that match the changed paths and connect fresh Evidence to each promise.
“Prepare this Task for closure. Run the required project checks and explain every missing, stale, or failed piece of Evidence.”
Skopos attaches proof to source and reports Readiness without optimistic guessing.
Keep Project Memory current
Review durable impact before finishing so the next agent learns the accepted result, not the old plan.
“Review whether this work changed architecture, decisions, patterns, or operating rules. Update only the canonical owner when needed.”
Skopos requires an attributed Memory review before Task closure.
Skopos supports the conversation. It does not replace it.
- Your coding agent still owns general reasoning and implementation.
- You still approve decisions that change direction, risk, or public behavior.
- Skopos protects project truth, Task intent, enforcement, recovery, and Evidence.