Pick 8 Tier

Input: $ARGUMENTS

Core Principles

1. Tiers reflect maturity, not value. Tier1 skills are the most tested and reliable. Tier4 skills are specialized or new. A tier4 skill can be exactly right for a specific task that no tier1 skill covers. Tier communicates confidence level, not importance.

2. Tier names have synonyms. Users will say “core” (tier1), “important” (tier2), “domain” (tier3), “specialized” (tier4), “experimental,” or “category.” The matcher must handle all of these.

3. Eight skills per tier shows the tier’s range. Each tier contains dozens of skills. Eight is enough to see the tier’s breadth without listing everything. The 8 should represent the tier’s diversity, not just its top-ranked skills.

4. Adjacent tier padding prevents empty results. Some tiers may have fewer than 8 skills. When this happens, pad from the adjacent tier (tier2 pads from tier1 or tier3) rather than returning a short list.

5. Tier context helps the user choose. The output should explain what this tier MEANS — what level of reliability, specialization, and depth the user can expect from skills in this tier.

Phase 1: Tier Mapping

[A] USER_INPUT: [from $ARGUMENTS]

[B] TIER_MATCH:

Step 1: Map input to tier:
    "core" | "tier1" | "essential" | "fundamental" → tier1
    "important" | "tier2" | "key" | "recommended" → tier2
    "domain" | "tier3" | "specialized domain" | "field-specific" → tier3
    "specialized" | "tier4" | "niche" | "specific" → tier4
    "category" | "router" | "orchestrator" → category
    "experimental" | "new" | "beta" | "untested" → experimental

Step 2: If no match: fuzzy match against tier descriptions
    tier1: "Core foundational skills — most tested, most broadly applicable"
    tier2: "Important skills — well-tested, frequently useful"
    tier3: "Domain-specific skills — deep expertise in particular areas"
    tier4: "Specialized skills — narrow focus, specific use cases"
    category: "Router skills — classify input and route to other skills"
    experimental: "Experimental skills — new, unproven, potentially valuable"

Step 3: If still no match: list available tiers and ask

    MATCHED_TIER: [tier name]
    MATCH_CONFIDENCE: [exact / synonym / fuzzy]

Phase 2: Tier Profile

[C] TIER_PROFILE:

    Tier: [name]
    Description: [what this tier means]
    Total skills in tier: [N]
    Reliability expectation: [high / medium / low / unknown]
    Typical use: [when you'd specifically want this tier]

    Key characteristics:
        - Depth: [typical line count range]
        - Connectivity: [avg invokes + invoked_by]
        - Category spread: [how many categories represented]

Phase 3: Selection

[D] POOL:

Step 1: Filter skills.json to MATCHED_TIER
    Pool size: [N]

Step 2: IF pool >= 8:
    Rank within tier by:
        CONNECTIVITY: invokes + invoked_by count
        DEPTH: line_count (higher = more developed)
        BREADTH: number of categories + tags
        COMPOSITE = CONNECTIVITY × 2 + DEPTH + BREADTH

    Diversity enforcement:
        - No more than 2 skills from any single category
        - Swap if needed: replace lowest-scoring duplicate-category with highest-scoring uncovered-category

    Take top 8 after diversity enforcement

Step 3: IF pool < 8:
    Take all [N] from matched tier
    Remaining = 8 - N
    Pad from adjacent tier:
        tier1 → pad from tier2
        tier2 → pad from tier1 or tier3
        tier3 → pad from tier2 or tier4
        tier4 → pad from tier3
        category → pad from tier1
        experimental → pad from tier4
    Mark padded skills as "[from adjacent tier: X]"

[E] SELECTED: [8 skills with scores and sources]

Phase 4: Tier Landscape

[F] LANDSCAPE:

Step 1: Group the 8 picks by category
    [category]: /[id], /[id]
    [category]: /[id]
    ...

Step 2: Identify tier superstars:
    TOP_IN_TIER: /[id] — [why this is the tier's best skill]

Step 3: Identify underappreciated skills:
    UNDERRATED: /[id] — [why this deserves more attention]

Step 4: Tier comparison:
    How does this tier compare to adjacent tiers?
    - vs [higher tier]: [what this tier lacks / what it has that higher doesn't]
    - vs [lower tier]: [what this tier offers / what lower tier specializes in]

Phase 5: Output

ALGORITHM: TIER
TIER: [matched tier]
MATCH: "[user input]" → "[tier name]" (confidence: [exact/synonym/fuzzy])
POOL: [N skills in tier]
PICKED: 8

TIER OVERVIEW:
  [tier name]: [1-2 sentence description]
  Reliability: [high/medium/low/unknown]
  Total skills: [N]
  Best for: [when you'd want skills from this tier]

8 SKILLS FROM [TIER]:

  1. /[id] — [title] ★ [if top-in-tier]
     [1-line description]
     Category: [cat] | Connections: [N] | Lines: [N]

  2. /[id] — [title]
     [1-line description]
     Category: [cat] | Connections: [N] | Lines: [N]

  [continue to 8...]

  [if any padded:]
  Note: Skills [N-8] are from adjacent tier [name] — included to fill the set

TIER LANDSCAPE:
  Categories represented: [list with counts]
  Top skill: /[id] — [why]
  Underrated: /[id] — [why]

TIER COMPARISON:
  vs [higher tier]: [key differences]
  vs [lower tier]: [key differences]

EXPLORE MORE:
  All [N] skills in [tier]: /pick [N] tier [tier name]
  Adjacent tier: /p8tier [adjacent tier name]
  Cross-tier best: /p10useful

Failure Modes

Depth Scaling

Default: 2x. These are floors.

Pre-Completion Checklist

• User input mapped to a tier with confidence level
• Tier described with overview, reliability, and use-case
• Exactly 8 skills returned
• No more than 2 per category within the 8
• Adjacent-tier padding clearly marked if used
• Top-in-tier and underrated skills identified
• Tier comparison with adjacent tiers provided
• Path to explore more of the tier shown

Integration

• Shortcut for: /pick 8 tier $ARGUMENTS
• Use when: You want to explore skills at a specific quality/maturity level
• Routes to: The 8 picked skills; /p8tier [adjacent] for comparison
• Related: /p7cat (filter by category instead), /p10useful (best across all tiers)
• Differs from /p7cat: cat filters by topic; tier filters by maturity level
• Differs from /p10useful: useful picks the best regardless of tier; tier constrains to one tier
• Differs from /p10diverse: diverse spans tiers and categories; tier focuses within one tier