FOWTDN - Figure Out What To Do Next
Input: $ARGUMENTS
Core Principles
-
“Next” means you’ve already started. This skill assumes prior work exists. If you’re at the very beginning with raw data and no direction, use
/fowtdfirst. This skill picks up mid-stream. -
What changed matters more than what exists. The right next action depends on what’s DIFFERENT since the last action — new information, resolved questions, shifted priorities, or surfaced problems. Start with the delta.
-
Momentum is real and valuable. If something is working, the default should be to continue it. Switching tracks has a cost. Only recommend a direction change if there’s a clear reason.
-
Diminishing returns are the main enemy. The most common failure is continuing to do what you were doing past the point where it’s producing value. Check for this explicitly.
-
One action, not a revised plan. The user wants to know what to do RIGHT NOW. Not a new roadmap. One move.
Phase 1: Progress Check
Where are you now vs. where you started?
[N1] ORIGINAL_GOAL: [what the user was trying to accomplish]
[N2] WORK_DONE: [what has been completed so far]
[N3] CURRENT_STATE: [what the user has/knows NOW that they didn't before]
[N4] PROGRESS_LEVEL: [early | mid | late | stuck | lost]
Progress Classification
| Level | Signals | Implication |
|---|---|---|
| Early | Just started, first pass done, initial triage complete | Next action should build on initial findings |
| Mid | Core work underway, some outputs produced, some gaps remaining | Next action should close the most important remaining gap |
| Late | Most work done, polishing, validating, or finalizing | Next action should be validation or completion |
| Stuck | Progress stalled, repeated attempts, circular work | Next action should break the loop — try a different approach |
| Lost | Goal unclear, direction confused, too many threads | Next action should be reorientation — /fowtd or /iagca |
Phase 2: Delta Assessment
What’s changed since the last action?
[N5] NEW_INFORMATION: [what you learned or received since last action]
[N6] RESOLVED_QUESTIONS: [uncertainties that are now settled]
[N7] NEW_QUESTIONS: [uncertainties that surfaced]
[N8] SHIFTED_PRIORITIES: [anything that's more or less important than before]
[N9] SURPRISES: [anything unexpected — positive or negative]
Delta Impact
[N10] DELTA_IMPACT: [confirms direction | suggests adjustment | demands pivot]
| Impact | Meaning | Response |
|---|---|---|
| Confirms | New info supports current direction | Continue, pick the natural next step |
| Adjusts | New info suggests a modification, not a change | Adjust and continue |
| Demands pivot | New info invalidates current approach | Stop current track, reassess |
If DEMANDS PIVOT:
[N11] PIVOT_REASON: [what specifically invalidated the current approach]
[N12] PIVOT_DIRECTION: [what the new direction should be]
Phase 3: Remaining Work Scan
What’s left to do?
[N13] REMAINING_ITEMS:
- [item] — VALUE: [high | medium | low] — EFFORT: [quick | moderate | heavy]
- [item] — VALUE: [value] — EFFORT: [effort]
- [item] — VALUE: [value] — EFFORT: [effort]
Value/Effort Matrix
| Quick | Moderate | Heavy | |
|---|---|---|---|
| High value | DO NEXT | DO SOON | PLAN CAREFULLY |
| Medium value | DO NEXT | MAYBE | PROBABLY NOT |
| Low value | ONLY IF FREE | SKIP | SKIP |
Phase 4: Next Action Selection
[N14] RECOMMENDED_NEXT: [one specific action]
[N15] WHY_THIS_NEXT: [why this action over the other remaining items]
[N16] WHAT_IT_PRODUCES: [concrete output or result]
[N17] DIMINISHING_RETURNS_CHECK: [is continuing the current track still the highest-value use of effort?]
Selection Rules
- If stuck: don’t do more of the same. Try a different angle, tool, or approach.
- If late stage: the next action should be finishing, validating, or shipping — not starting something new.
- If new information surfaced: address it before continuing the old plan.
- If priorities shifted: re-rank remaining work before picking.
- If everything remaining is low-value: the next action might be “stop — you’re done enough.”
Phase 5: Output
FIGURE OUT WHAT TO DO NEXT
==========================
STARTED WITH: [original goal]
YOU'VE DONE: [brief summary of progress]
YOU'RE AT: [progress level]
WHAT CHANGED:
[most important delta — new info, resolved question, or surprise]
REMAINING WORK: [count of items, with high-value count]
DIMINISHING RETURNS? [yes/no — is continued effort still high-value?]
→ DO THIS NEXT: [one specific action]
WHY: [one sentence — why this over alternatives]
PRODUCES: [what you'll have after doing it]
SKILL: → INVOKE: /skill-id $ARGUMENTS
AFTER THAT:
IF_GOOD: [likely next step if this works]
IF_STUCK: [what to try if this doesn't produce results]
IF_DONE: [how to know you can stop]
Failure Modes
| Failure | Signal | Fix |
|---|---|---|
| Continuing past diminishing returns | Recommending more of the same when value is declining | Explicitly check: is this still the highest-value use of effort? |
| Ignoring the delta | Recommending the “planned” next step when new info changes things | Phase 2 must be done before Phase 4 — new info can change everything |
| Restarting instead of continuing | Sending user back to /fowtd when they have momentum | Only reorient if genuinely LOST, not just slightly uncertain |
| Too many next steps | ”Do A, then B, then C” | ONE action. Route to /fonss if they need a sequence |
| Not knowing when to stop | Never outputs “you’re done” | Check remaining work — if all low-value, recommend stopping |
| Forward-only bias | Never recommends revisiting or backtracking | Sometimes the best next step is to redo or reconsider a prior step |
Depth Scaling
| Depth | Progress Check | Delta Analysis | Remaining Scan | Alternative Actions |
|---|---|---|---|---|
| 1x | Level only | Most important change | Top 3 items | 1 action |
| 2x | Level + evidence | Full delta | All items ranked | 2-3 actions, pick best |
| 4x | Full + trajectory | Delta + impact analysis | All items with value/effort | 4-5 actions, scored |
| 8x | Full + pattern check | Delta + second-order effects | Comprehensive with dependencies | Full action comparison |
Default: 2x. These are floors.
Pre-Completion Checklist
- Progress level assessed (not assumed)
- Delta identified — what’s new since last action
- If delta demands pivot: pivot direction specified
- Remaining work scanned and ranked
- Diminishing returns explicitly checked
- ONE next action selected (not a sequence)
- Action is concrete and specific
- Recovery path specified (IF_STUCK)
- Completion condition specified (IF_DONE — how to know when to stop)
- Backtracking considered as valid option
Integration
- Use from: user is mid-process and needs direction on what to do next
- Differs from
/fowtd: fowtd starts from raw data with no direction; fowtdn picks up mid-stream - Differs from
/next: next is a general next-step router; fowtdn is specifically for data/work-in-progress - Differs from
/wtdn: wtdn is simpler — situation/goal/gap/move; fowtdn adds delta analysis and diminishing returns - Differs from
/fonss: fonss builds a multi-step chain; fowtdn picks ONE next action - Routes to: whatever skill the next action requires
- Falls back to:
/fowtdif user is actually lost and needs full reorientation