Pick 10 Random

Input: $ARGUMENTS

Core Principles

Random is a feature, not a bug. Optimization algorithms surface the same high-tier skills repeatedly. Random selection breaks this pattern and surfaces skills the user would never encounter through goal-directed search.
No filtering, no weighting. Every skill in the library has an equal probability of selection. Tier1 and tier4, well-known and obscure — all equally likely. This is the defining property. If you weight by anything, it’s not random.
The pool can be constrained. While selection within the pool is uniform, the pool itself can be narrowed. /p10random tier3 or /p10random Research constrains the pool first, then picks randomly within it. The constraint is explicit, not hidden.
Random selections demand context. Because the picks are not curated for a purpose, each one needs enough description for the user to understand why they might care. Random without context is just noise.
Re-rolling is expected. Random is low-commitment. If the 10 picks don’t interest the user, running it again costs nothing and produces a completely different set. This is a feature of the algorithm.

Phase 1: Pool Definition

[A] POOL_CONSTRAINT: [parse from $ARGUMENTS — if any]

Step 1: Check if $ARGUMENTS contains a constraint:
    - Tier name (e.g., "tier3", "core", "experimental") → filter to that tier
    - Category name (e.g., "Research", "Decision") → filter to that category
    - No constraint → use full library

Step 2: Load the pool
    POOL_SOURCE: [full library / tier: X / category: X]
    POOL_SIZE: [N skills in pool]

Step 3: Verify pool is large enough
    IF pool < 10: → WARNING: Pool has only [N] skills. Returning all [N].
    IF pool >= 10: → proceed with random selection

Phase 2: Random Selection

[B] SELECTION:

Step 1: Assign each skill in the pool a random position
Step 2: Take the first 10 (or all if pool < 10)
Step 3: No re-ranking, no re-ordering by quality — preserve the random order
    (This communicates that the ordering is arbitrary, not a ranking)

[C] SELECTED:
    1. [skill ID]
    2. [skill ID]
    ...
    10. [skill ID]

Phase 3: Context Generation

For each selected skill, provide enough context for the user to decide if it interests them.

[D] CONTEXTUALIZED:

For each skill:
    Step 1: Read its metadata from skills.json (title, description, tier, categories, tags)
    Step 2: Summarize in one line what it DOES (not what it IS)
    Step 3: Note its tier and primary category
    Step 4: Flag if it's particularly unusual or specialized (surprise value)

Phase 4: Post-Selection Analysis

[E] DISTRIBUTION_ANALYSIS:

    Categories represented: [list with counts]
    Tiers represented: [list with counts]
    Functions represented: [analysis/decision/planning/creation/validation — which are present]

    ACCIDENTAL_COVERAGE: [did random selection happen to cover anything well?]
    NOTABLE_CLUSTERS: [did random selection cluster in any area?]
    SURPRISE_PICKS: [which of the 10 are least commonly seen in curated lists?]

Phase 5: Output

ALGORITHM: RANDOM
POOL: [pool size] ([full library / constrained to: X])
PICKED: 10

  1. /[id] — [title]
     [1-line: what it does]
     Tier: [tier] | Category: [category]

  2. /[id] — [title]
     [1-line: what it does]
     Tier: [tier] | Category: [category]

  [continue to 10...]

DISTRIBUTION:
  Categories: [list]
  Tiers: [list]
  Functions: [list]

SURPRISE PICKS: [which of these are least commonly encountered]

NOTE: This is a random selection — not a recommendation.
Re-run /p10random for a completely different set.
For curated picks, try /p10goal [your goal] or /p10useful.

Failure Modes

Failure	Signal	Fix
Hidden weighting	Selection clearly biased toward high-tier or popular skills	Verify: every skill in pool had equal chance. Do not “improve” random
No context	Skill listed with just ID and title — user can’t evaluate	Provide 1-line action description for each pick
Treating random as recommendation	Output implies these are the “right” skills	Explicitly state “this is random, not curated”
Pool constraint not honored	User said “random tier3” but tier1 skills appear	Verify pool was filtered before selection
Re-ranking after selection	Picks are subtly re-ordered by quality	Preserve random order — do not sort by tier or score
Insufficient pool warning	Pool has 5 skills but output says “picked 10”	If pool < N, warn and return all available

Depth Scaling

Depth	Pool	Selection	Context
1x	Full library, no analysis	10 random picks	ID + title + tier
2x	Pool + constraint handling	10 random + distribution analysis	1-line descriptions + surprise flags
4x	Pool analysis + size verification	10 random + full distribution + coverage check	Descriptions + connections + suggested exploration order
8x	Pool comparison across tiers/categories	Multiple random draws + statistical coverage	Full skill summaries + comparison to curated alternatives

Default: 2x. These are floors.

Pre-Completion Checklist

Pool constraint parsed and applied (if any)
Selection is genuinely random (no hidden weighting)
Exactly 10 skills returned (or all if pool < 10)
Each pick has a 1-line action description
Distribution analysis shows what categories/tiers/functions are represented
Output explicitly states this is random, not curated
Suggestion provided for curated alternatives

Integration

Shortcut for: /pick 10 random
Use when: You want serendipitous discovery, not targeted search
Routes to: The 10 picked skills; re-run for different set
Related: /p10diverse (systematic coverage), /p10useful (quality-ranked)
Differs from /p10diverse: diverse guarantees coverage; random does not
Differs from /p10useful: useful ranks by quality; random ignores quality
Variant: /p10random [tier/category] constrains the pool before random selection

p10random - Pick 10 Random

Pick 10 Random

Core Principles

Phase 1: Pool Definition

Phase 2: Random Selection

Phase 3: Context Generation

Phase 4: Post-Selection Analysis

Phase 5: Output

Failure Modes

Depth Scaling

Pre-Completion Checklist

Integration

`p10random - Pick 10 Random`