| name | workflow-state |
| description | Create and update .cursor/plans/workflow_state.md at each ks-conductor phase boundary. Provides a durable phase-tracking file that survives context compaction, long autonomous runs, and session interruptions. Use alongside /ks-conductor for STANDARD and COMPLEX tasks.
|
When to use
- Running a STANDARD or COMPLEX
/ks-conductor task
- At any phase boundary (INTAKE → SCOUT → PLAN → BUILD → VERIFY → CLOSE)
- Resuming after context compaction or session interruption
- A long autonomous run needs a durable checkpoint on disk
On-demand loading: Read this skill only when the task clearly matches the description above. Do not load for unrelated work.
workflow-state
Purpose
Maintain a single .cursor/plans/workflow_state.md file that tracks the
current ks-conductor run state — phase, status, plan steps, and log entries.
This file survives context compaction and long autonomous runs, letting the
agent resume from a known checkpoint even after context loss.
Complements (does not replace) the existing .cursor/plans/ SDLC artifacts
(intake-brief-.md, scout-.md, execution-pack-*.md, etc.).
When to activate
- At the start of every STANDARD or COMPLEX ks-conductor run
- At each phase boundary to update status and log
- When the agent detects context was compacted (
precompact-preserve.js fired)
- When the user asks "where did we leave off?" or "what phase are we in?"
- When resuming a run after interruption
File location
Always write to: .cursor/plans/workflow_state.md in the active project workspace (not in ~/.cursor).
If no project workspace is active (dotfiles workspace), write to ~/.cursor/plans/workflow_state.md.
See references/workflow-state-template.md for the full file template and per-phase update instructions.
Phase boundary updates (summary)
The update must be atomic — write the whole file, not a partial append. Always update "Last updated" on every write.
| Phase end | Set Phase → | Key action |
|---|
| INTAKE | SCOUT | Fill task summary, run ID, user-devcontext conversation ID |
| SCOUT | PLAN or BUILD | Fill classification, workflow type, artifact paths |
| PLAN | CRITIC or BUILD | Fill Plan steps from execution pack |
| BUILD start | BUILD | Set "Current focus" to first task |
| BUILD checkpoint | BUILD | Check off step ✓, append log, update focus |
| VERIFY | CLOSE or FIX_LOOP | Log GREEN/YELLOW/RED result |
| BLOCKED | — | Set Status BLOCKED, fill Blockers field |
| CLOSE | CLOSE | Set Status COMPLETED, log final outcome |
Resume protocol
When the agent detects possible context loss or the user asks to resume:
- Read
.cursor/plans/workflow_state.md
- Read the referenced artifact files (scout, execution-pack, etc.)
- Announce: "Resuming from [Phase] — [Status]. Last log entry: [entry]."
- Continue from current Phase without re-doing completed phases
- Update the state file immediately on resuming
Integration notes
- Do NOT replace existing SDLC artifacts (
intake-brief-*.md, scout-*.md, etc.) — workflow_state.md is a lightweight companion
- Do NOT store large content in workflow_state.md — keep each field to 1-3 lines; link to artifact files for details
- Always update "Last updated" timestamp on every write
- The file is expected to be overwritten (not appended) on each update
