Assessing Methodology Completeness

Progress:

• Step 1: Parse input for the five methodology signals
• Step 2: Score each signal present/absent/partial, aggregate to 0-100
• Step 3: If score ≥ 60, pass input directly to generation
• Step 4: If score < 60, write reasoning (1-2 sentences) + missing elements list
• Step 5: Present 3 extraction questions targeting the weakest signals
• Step 6: If answered, merge answers into original input and generate; if skipped, generate with original input as-is

Step 1-2: Scoring the Five Signals

Evaluate each signal independently, then combine:

Score each signal as: absent (0), vague/implied (10), explicit (20). Sum for total 0-100.

Step 3: Fast Path (Score ≥ 60)

Generate the skill directly from the original input. No follow-up needed.

Step 4-6: Extraction Path (Score < 60)

Write reasoning as one sentence naming what's present and one naming what's missing. Example: "Your input describes what you do but not how you do it or what results it produces."

List 2-4 missing elements as short, specific phrases (not generic advice)—each should map to a concrete gap in the five signals, e.g. "a numbered sequence of steps" rather than "more detail."

Generate exactly 3 questions, each targeting the lowest-scoring signals. Prioritize signals scored 0 over signals scored 10.

• Question templates by missing signal:
  • Missing process → "Walk me through your process step-by-step, from start to finish."
  • Missing tools → "What specific tools, software, or frameworks do you use?"
  • Missing examples → "Can you give a concrete example: what input did you start with, and what output did you produce?"
  • Missing outcomes → "What measurable results have you achieved (numbers, %, time saved)?"
  • Missing decision logic → "Are there situations where you'd do something differently? What triggers the change in approach?"

If user answers all 3, merge answers into the original input as additional context, then generate. If user skips, generate with the original input unmodified—do not block.

Example 1: Input: "I review pull requests for security issues." Output:

Score: 25
Reasoning: States the activity but gives no process, tools, or measurable results.
Missing elements: specific vulnerability checklist, tools used (e.g. static analyzers), example of a caught issue
Questions:
1. Walk me through your review process step-by-step.
2. What tools or checklists do you use to spot security issues?
3. Can you share an example: what vulnerable code looked like, and what you flagged?

Example 2: Input: "I use RFM analysis (Recency, Frequency, Monetary) to segment customers into 5 tiers, then trigger tier-specific email campaigns via Klaviyo. Tier 1 customers get a 10% discount if they haven't purchased in 30 days; otherwise they get a loyalty reward. This has increased repeat purchase rate by 22% on average across 6 clients." Output:

Score: 95
Reasoning: N/A — proceeds directly to generation.
Action: Generate skill immediately.