mbwsims/claude-universe/forecast

Forecast

Predict which files are about to become problems. Hotspots look backward — what's risky now. Forecast looks forward — what's BECOMING risky. By the time a file is a hotspot, it's already painful. Forecast catches it while there's still time to intervene.

Workflow

1. Scope

• No argument: Analyze the whole project. Surface the top 5 files on concerning

trajectories.

• With argument: Focus on a specific directory.

2. Gather Trend Data

With timewarp-mcp (preferred): Call timewarp_trends to get computed growth rates, acceleration data, and churn trends for all source files. This provides precise time-series data sampled at intervals. Then call timewarp_history on the top concerning files to add commit-count and author-context before writing recommendations.

Without timewarp-mcp: Compute manually by sampling git history:

# Line count 6 months ago vs now
git log --oneline --since="6 months ago" --until="3 months ago" --format=format: --name-only | sort | uniq -c | sort -rn | head -20
git log --oneline --since="3 months ago" --format=format: --name-only | sort | uniq -c | sort -rn | head -20

Compare churn in the first half vs second half of the period to detect acceleration.

3. Analyze Trends

For each source file with meaningful history, assess:

• Churn trend: Is it changing more frequently recently than historically?
• Accelerating: more commits per month now than 3 months ago
• Stable: consistent commit rate
• Decelerating: activity is tapering off
• Complexity growth: Is it getting larger/more complex?
• Track: line count, function count, import count over time
• A file growing 10% per month will double in 7 months
• Author fragmentation: Are more people touching it? (knowledge spreading thin)

See references/trend-analysis.md for methodology.

4. Project Forward

For files on concerning trajectories, project the trend forward:

• When will it cross size thresholds? (300-500 lines = large/hard to reason about, >500 lines = critical/should be split)
• When will churn make it a hotspot? (if not already)
• Is the growth accelerating (exponential) or linear?

5. Check for Cached Data

Read .timewarp/ for existing drift or dissect data on trending files. If a file is both drifting AND growing, that's a stronger signal — note the connection.

6. Present and Save

Report format:

## Forecast — {scope}
 
Analyzed {n} source files over {period}
 
### Files on Concerning Trajectories
 
| Rank | File | Growth Pattern | Lines/Month | %/Month | Churn Pattern | Projection (3mo) | Projection (6mo) | Threshold |
|------|------|---------------|-------------|---------|--------------|------------------|------------------|-----------|
| 1 | src/lib/auth.ts | accelerating | +26.7 | +14.8% | accelerating | ~425 lines | ~560 lines | 500 in ~5mo |
| 2 | src/api/orders.ts | linear | +15.0 | +8.2% | accelerating | ~345 lines | ~390 lines | 500 in ~11mo |
| 3 | src/lib/db.ts | linear | +8.3 | +5.0% | linear | ~225 lines | ~250 lines | 300 in ~9mo |
 
These columns map directly to `timewarp_trends` output: `growth.pattern`, `growth.linesPerMonth`, `growth.percentPerMonth`, `churn.pattern`, `projection.linesIn3Months`, `projection.linesIn6Months`, `projection.crossesThreshold`.
 
### Detailed Analysis
 
**#1: src/lib/auth.ts** — Complexity Accelerating
- Sample trend: 180 lines / 8 functions → 240 lines / 11 functions → 310 lines / 15 functions
- Growth is accelerating — added more in the last 3 months than the previous 3
- Recommendation: Split before it crosses 400 lines. Auth verification, session
  management, and token handling could be separate modules.
 
**#2: src/api/orders.ts** — Churn Accelerating
- ...
 
### Stable Files (No Concern)
{Brief note on files that are stable or decelerating — positive signal}
 
### Recommendations
{Prioritized list: which files to address first and what to do}

Save results to .timewarp/forecast-{date}.json with structured trend data that other skills can consume.

`.timewarp/` directory: Create the directory if it doesn't exist. Results older than

30 days are stale — prefer re-running the analysis over consuming old data. Other

Timewarp skills may read these files to cross-reference findings (e.g., /forecast

checks for drift data on trending files).

Guidelines

• Focus on ACCELERATION, not just size. A 500-line file that hasn't changed in a year is

fine. A 200-line file growing 15% per month is a forecast concern.

• Don't cry wolf. Only surface files with genuinely concerning trajectories. If nothing

is trending badly, say "no concerning trajectories found" — that's a good result.

• Projections are estimates, not predictions. Use language like "at current rate" and

"approximately." Never give false precision.

• Connect to action. Every forecast entry should include a specific recommendation —

what to do about it and when.

• If git history is short (< 3 months), note that projections are less reliable and

focus on current growth rate rather than acceleration.

Related Skills

• `/drift` — Check if forecasted files are also drifting from their purpose
• lenskit `/hotspots` — Current risk assessment (forecast's backward-looking complement)

Additional Resources

• `references/trend-analysis.md` — Growth rate computation, acceleration detection,

threshold calibration, false alarm avoidance