| name | complete-project |
| description | This skill should be used when the user asks to 'complete a project', 'wrap up a project', 'archive a project', or 'run project completion'. |
Project Completion Orchestrator
Guides a project through its full completion lifecycle: distillation, memory review, system doc proposals, and archival. Each phase pauses for user confirmation so nothing is irreversible without explicit approval.
MCP Tools
listWorkspaces — resolve the active workspace and obtain its workspaceId before calling listProjects.
getProjectDetails — fetch project metadata, status, and workspace context.
listProjects — find the active project if not provided.
startProjectCompletion — set status to completing and return ranked candidate fleeting memories for distillation review.
archiveProject — mark project as archived and remove agent sessions. Plan, tasks, and promoted memories are preserved.
Phase 1: Resolve Project
- Identify the current project from session context (active project name/slug or the user's explicit mention).
- Call
getProjectDetails({ projectId }) to confirm the project exists and retrieve its workspace slug and status.
- Branch on status:
- If
status is archived: print "Project is already archived; nothing to do." and stop.
- If
status is completing: ask the user "Project is mid-completion. Resume from Phase 3 (review memories) or restart from Phase 2 (distillation)?" and proceed based on their answer.
- Otherwise: print the project name and current status, then continue to Phase 2.
Phase 2: Distillation
- Call
startProjectCompletion({ projectId }).
- This sets the project status to
completing.
- It returns
{ candidates: FleetingMemory[] } — unpromoted fleeting memories scoped to the workspace, ranked by signal score.
- Print the count of candidate fleetings (e.g., "Found 12 candidate memories for review").
- Pause. Present the candidates briefly and ask the user: "Ready to proceed to memory review? This will hand off to
/engy:review-memories."
- Wait for explicit user confirmation before continuing.
Phase 3: Memory Review
After user confirms, invoke engy:review-memories directly via the Skill tool:
Skill({ skill: 'engy:review-memories' })
Skill(...) runs the inner skill to completion and returns control here before Phase 4 begins; do not advance until it returns. The invoked skill re-resolves workspace/project context itself — surface the active project name and workspace slug to the user before invoking so they can confirm the inner skill will target the right scope. The review skill is interactive — the user works through each candidate (approve, edit, supersede, contradict, skip). Do not rush or batch-skip. When it returns, continue to Phase 4.
Phase 4: System Doc Proposals
Pause and ask: "Memory review is complete. Generate system doc proposals from the project's completed tasks and promoted memories? [yes/skip]"
Phase 5: Archive
- Pause. Confirm with the user: "Memory review and doc proposals are complete. Archive the project now? This will mark it archived and preserve all tasks, plan content, and promoted memories."
- On confirmation, call
archiveProject({ projectId }).
- Print confirmation: "Project archived. Agent sessions and execution logs have been removed. Plan, tasks, and memories are preserved."
Key Principles
- Pause between phases. Never run phases back-to-back without user confirmation. Each pause is a checkpoint where the user can stop, inspect, or continue later.
- Phases are resumable. If the user stops mid-flow, they can re-invoke
/engy:complete-project and pick up from where they left off by checking the project's current status via getProjectDetails.
- Archival is final. Make sure the user understands archival removes agent sessions and execution logs before calling
archiveProject.
Flow Position
Previous: Active project with tasks in done status | Next: Knowledge available via workspace-level memory for future projects