search-optimizer

name	search-optimizer
description	Recommend job search strategy adjustments based on application data. Use this skill when the user asks "what should I apply to next?", "where should I focus?", "how should I adjust my search?", "what types of roles are working?", "should I keep doing stretch applications?", "am I wasting time on LinkedIn?", "what's my best approach?", or any strategic question about job search direction. Also trigger for weekly check-ins, after a batch of rejections, or when the user seems discouraged and needs data-driven guidance. This skill interprets analytics data and makes specific recommendations — it doesn't just show numbers (that's application-analytics). Do NOT trigger for individual application evaluation (job-intake) or status updates (job-tracker).
recommended_model	{"default":"opus","reasoning":"Strategic career reasoning requires weighing multiple factors: market signals, personal strengths, conversion data, and career trajectory. Opus handles the nuanced trade-offs between \"apply to more stretch roles\" vs \"focus on strong matches\" based on real data.\n","downgrade_to_sonnet_when":"The user wants a simple recommendation like \"which source to use\" without deeper strategic analysis.\n"}

Search Optimizer — Data-Driven Strategy

Intent

1. User preferences are constraints, not suggestions — "only remote" or "won't go below Director" are boundaries to optimize within, not objections to overcome
2. Career goals guide interpretation of data — the same conversion rate means different things depending on whether the candidate is pivoting industries or climbing within one
3. Empathy in framing — job searching is stressful; frame recommendations as optimization, not criticism; "your referrals convert 3x better" is actionable, "you're wasting time on LinkedIn" is demoralizing
4. Data-backed claims only — every recommendation traces to a specific number or pattern; gut feelings belong to the candidate, not the advisor
5. Strategic clarity over information overload — 3-5 specific recommendations with evidence beats a 20-point listicle
6. Burnout prevention through focus — concentration of effort on high-converting channels prevents the spray-and-pray exhaustion that kills job searches
7. Meaningful recommendations require 5+ outcomes; 3-5 adjustments per review; trigger monthly or per 5 new outcomes — frequency and depth match the data available

Why this skill exists

Without strategy, job searching is random. You apply everywhere, burn out, and wonder why nothing's working. This skill looks at your data and says: "Your referrals convert 3x better than LinkedIn. Your strong matches convert 70%. You've been rejected from 3 stretch roles for the same reason. Here's what to change."

Prerequisites

• job-applications/analytics-report.md — Run application-analytics first if this doesn't exist (or data is stale)
• At least 5+ applications with outcomes for meaningful recommendations
• master/narrative.md — For understanding the user's positioning and goals

With fewer than 5 applications, provide directional guidance based on available data rather than definitive strategy.

Workflow

Step 1: Load the data

Read:

• analytics-report.md (or scan metadata.json files if report is stale/missing)
• master/narrative.md for the user's career themes and targets
• All interview-notes.md files for market signals
• All learning_flags across metadata.json files

Step 2: Assess current strategy

Source effectiveness:

• Which sources are producing interviews?
• Which sources have the best time-to-response?
• Is the user over-invested in a low-converting source?

Match score strategy:

• What's the conversion rate by match score?
• Are stretch applications ever working, or just burning time?
• Are strong matches converting reliably?

Role type fit:

• Which role titles and seniority levels are converting?
• Is the user's positioning (from narrative.md) aligned with what's working?
• Are certain industries or company sizes working better?

Timing and velocity:

• How many applications per week?
• Is the user applying fast enough to maintain pipeline?
• Is there a sweet spot for follow-up timing?

Step 3: Identify strategic adjustments

Based on the data, develop 3-5 specific recommendations:

Source reallocation:

If referral conversion > 2x LinkedIn conversion:
  → "Shift 60% of effort to building referral pipeline"
  → Specific actions: reach out to X, Y, Z contacts from achievements.md companies

Match score focus:

If strong matches convert > 50% and stretch matches convert < 10%:
  → "Focus on strong and good matches. Stretch applications only if the company
     is a top-5 target worth the investment."
If stretch matches are converting:
  → "Your positioning is working for stretch roles. Keep the mix."

Positioning adjustments:

If 3+ rejections cite the same reason:
  → "Your resume is hiding [X]. Move it above the fold."
If interview debriefs show recurring topic:
  → "Market is asking for [X]. Make it prominent in your summary."

Pipeline health:

If fewer than 3 applications in "applied" or "interviewing":
  → "Pipeline is thin. Target 2-3 new applications this week."
If more than 10 applications with no interviews:
  → "Something fundamental needs to change. Consider: resume restructure,
     different role targets, or different sources."

Step 4: Produce the strategy document

Write job-applications/search-strategy.md:

# Search Strategy — [Date]

## Current Position
[1-2 sentences summarizing where the search stands]

## What's Working
- [Data-backed observation]
- [Data-backed observation]

## What's Not Working
- [Data-backed observation with specific evidence]

## Recommendations

### 1. [Primary recommendation]
**Why:** [Evidence from data]
**Action:** [Specific next step]

### 2. [Secondary recommendation]
**Why:** [Evidence from data]
**Action:** [Specific next step]

### 3. [Third recommendation]
**Why:** [Evidence from data]
**Action:** [Specific next step]

## Target Profile for Next Applications
- **Role types:** [specific titles that work]
- **Company types:** [industry, size, stage]
- **Sources to prioritize:** [ranked by conversion]
- **Match score filter:** [minimum viable match]
- **Keywords to emphasize:** [from interview patterns]

Step 5: Present and discuss

Share the strategy with the user. Be direct but empathetic — job searching is stressful. Frame recommendations as optimization, not criticism.

If the data contradicts the user's instincts, explain why with specific numbers. "I know LinkedIn feels productive because you're active there daily, but 0 of your 6 LinkedIn applications have led to interviews. Your 2 referral applications both led to interviews."

Edge cases

• User has strong preferences — Respect stated constraints ("I only want remote" or "I won't go below Director level"). Optimize within those bounds.
• Not enough data — Be honest. "With 4 applications, I can see early signals but not reliable patterns. Here's what the early data suggests..."
• Everything is working — Great! Recommend maintaining the approach and increasing volume.
• Nothing is working — The hardest case. Look for the root cause: wrong role level, wrong industry, resume not showcasing key skills, or market timing.

Skill composition

Upstream	This skill	Downstream
application-analytics (data)	search-optimizer (strategy)	job-intake (informed by strategy)
interview-debrief (signals)		resume-tailor (adjusted approach)
application-outcome-logger (patterns)

Read references/optimization-patterns.md for strategy frameworks.