agenticnotetaking/skills/plan-continue

/eidos:plan-continue

Resume work on an existing plan file.

Usage

/eidos:plan-continue [plan-name]

Instructions

1. Find Active Plans

Search memory/ for plan - *.md files. Identify incomplete plans by checking:

• Actions without [x] completion markers (note: [p] postponed actions don't count as incomplete — a plan with only [x] and [p] actions may be effectively done)
• Status field not completed or abandoned
• Progress Log entries indicating in-progress work

2. Select Plan

Single incomplete plan:

Only one incomplete plan found: [[plan - 2602101030 - add user auth]]

Next action: Implement login endpoint

continue?

Multiple incomplete plans:

Active plans found:

1 - [[plan - 2602101030 - add user auth]]
  - 3/5 actions complete
  - Next: Implement login endpoint

2 - [[plan - 2602091500 - refactor database layer]]
  - 1/4 actions complete
  - Next: Add connection pooling

Which plan to continue?

No incomplete plans:

No incomplete plans found.

Options:
1 - Create a new plan with /eidos:plan
2 - View completed plans

3. Load Context

When a plan is selected, read in order:

1. The plan template — [[template - plan - structured phases with actions and progress tracking]] (understand expected structure)
2. The plan file — understand actions and progress
3. Linked specs in eidos/ — refresh on requirements
4. Linked goals in memory/ — understand desired outcome
5. Observations — check for open (non-FIXED) observations that need implementation
6. Progress Log — see where we left off
7. Learnings — apply discovered insights

4. Present Current State

## Resuming: [[plan - 2602101030 - add user auth]]

**Spec:** [[spec - user authentication]]

### Progress
1. [x] Set up auth middleware
   - => chose Express middleware over custom wrapper
2. [x] Create user model
   - => [[spec - user model - email based identity]] created
   - => found existing password hashing in utils/crypto.ts
3. [ ] Implement login endpoint
4. [ ] Implement logout endpoint
5. [p] Add session management
   - => deferred — cookie-based auth sufficient for now

### Next Action
Implement login endpoint (skipping 5 — postponed)

Ready to continue?

5. Continue Work

When user confirms:

1. Begin work on next incomplete [ ] action (skip [p] postponed actions)
2. Commit after each logical change
3. IMPORTANT: After every completed action, update the plan file and commit it.
   • Before marking [x], verify every sub-bullet against actual output (run it, look at it, test it)
   • Only mark [x] when all sub-bullets are satisfied — partial stays [ ] with => notes for what's done and what remains
   • Add => sub-notes that account for every sub-bullet, not just what was built
   • Add brief Progress Log entry with timestamp
   • The plan is the persistent record — chat is ephemeral.

Action loop:

implement → verify each sub-bullet → commit code → update plan ([x] or [ ] with => notes, Progress Log) → commit plan → continue?

The plan is best-effort. If an action turns out wrong, underspecified, or based on a bad assumption, don't grind through it — surface the issue, propose a plan edit or concrete options, and let the human steer. When the gap is missing knowledge, suggest /eidos:research; when it's unclear design, suggest /eidos:spec. Record the change in the plan's Adjustments. Correcting a plan mid-flight is expected, not a deviation.

6. Quick Continuation

After completing each action:

Completed: Implement login endpoint

Next: Implement logout endpoint
continue?

When a phase completes and the result is something the user can see or try, tell them what's now testable:

Phase 2 complete.

You can now test: [what to try, where to look, what command to run]

Next: Phase 3 — [description]
continue?

7. Handling Observations

When the user reports issues during testing (what they see vs what they want), use /eidos:observe to structure them. Observations become numbered entries in the plan and generate implementation tasks.

When resuming a plan that has open (non-FIXED) observations:

• Note them in the state presentation
• Observation-generated tasks are normal actions — pick them up in order like any other
• When an observation's tasks are all done, mark it FIXED

Output

• Resumes work on selected plan
• Updates plan file with progress after each action

When to Use

• Starting a new session to continue previous work
• Returning to a paused plan
• Checking status of active work