| name | p10useful - Pick 10 Useful |
| description | Pick the 10 most generally useful skills based on tier, connectivity, breadth, and depth. The 'greatest hits' of the skill library. Standalone implementation of the USEFUL algorithm from /pick. |
| tier | tier4 |
| categories | ["Meta & Navigation"] |
| tags | ["shortcut","pick","discovery","useful","quality"] |
| output | {"format":"prose"} |
Pick 10 Useful
Input: $ARGUMENTS
Core Principles
-
Usefulness is breadth times depth. A skill is useful if it applies to many situations (breadth) AND produces substantial output when it does (depth). High breadth + low depth = superficial. Low breadth + high depth = niche. High both = useful.
-
Connectivity is a signal. Skills that invoke many other skills or are invoked by many other skills sit at network hubs. Hub skills are useful because they are entry points to larger capability chains.
-
Tier reflects community validation. Higher-tier skills have survived more usage and refinement. Tier is not a perfect proxy for usefulness, but it is a strong prior.
-
Category diversity prevents echo chambers. The 10 most useful skills should not all be analysis skills. Cap any single category at 3 to ensure the user gets a well-rounded toolkit.
-
The list should be defensible. For each pick, you should be able to answer: "If a new user could only learn 10 skills, why THIS one?" If you can't answer that, the pick is not in the top 10.
Phase 1: Universal Scoring
Score every skill in the library on four dimensions.
[A] SCORING_DIMENSIONS:
For each skill:
1. TIER_SCORE:
tier1 = 10
tier2 = 8
category = 7
experimental = 6
tier3 = 4
tier4 = 2
2. CONNECTIVITY:
Count of (invokes + invoked_by)
Skills with 0 connections = 0
Skills with 1-3 connections = 2
Skills with 4-7 connections = 4
Skills with 8+ connections = 6
3. BREADTH:
Count of non-empty categories + count of tags
0-2 = 1
3-5 = 2
6-8 = 3
9+ = 4
4. DEPTH_PROXY:
line_count > 200 = 4
line_count 100-200 = 3
line_count 50-100 = 2
line_count < 50 = 1
TOTAL = TIER_SCORE + CONNECTIVITY + BREADTH + DEPTH_PROXY
Phase 2: Ranking and Diversity Enforcement
[B] RANKING:
Step 1: Sort all skills by TOTAL score (descending)
Step 2: Take top 10
Step 3: DIVERSITY CHECK:
Count skills per category in the top 10
IF any category has > 3 skills:
→ Remove the lowest-scoring excess skills from that category
→ Replace with the highest-scoring skills from underrepresented categories
Repeat until no category has > 3 skills
Step 4: FUNCTION CHECK:
Verify at least 4 of these functions are represented:
[analysis / decision / planning / creation / validation / exploration / diagnosis]
IF < 4:
→ Swap lowest-scoring pick for highest-scoring from missing function
Phase 3: Justification
For each of the 10 picks, generate a defensible justification.
[C] JUSTIFICATIONS:
For each pick:
QUESTION: "If a new user could only learn 10 skills, why this one?"
ANSWER: [1-2 sentence justification]
SCORE_BREAKDOWN: tier=[X] + connectivity=[Y] + breadth=[Z] + depth=[W] = [total]
UNIQUE_VALUE: [what does this skill do that no other top-10 skill does?]
Phase 4: Output
ALGORITHM: USEFUL
POOL: [total skills scored]
PICKED: 10
TOP 10 MOST USEFUL SKILLS:
1. /[id] — [title] — Score: [X] (tier=[T], connections=[C], breadth=[B], depth=[D])
[1-line description]
Why top 10: [justification]
2. /[id] — [title] — Score: [X] (tier=[T], connections=[C], breadth=[B], depth=[D])
[1-line description]
Why top 10: [justification]
[continue to 10...]
CATEGORY DISTRIBUTION:
[category]: [N picks]
[category]: [N picks]
...
FUNCTION COVERAGE:
[analysis/decision/planning/creation/validation/exploration/diagnosis]: [which are covered]
HONORABLE MENTIONS (skills 11-15):
11. /[id] — [title] — Score: [X]
12. ...
NOTABLE ABSENCES:
[any skill type or category conspicuously missing from the top 10]
To fill: /[suggested skill]
Failure Modes
| Failure | Signal | Fix |
|---|
| Category clustering | 5+ picks from analysis or decision skills | Enforce 3-per-category cap |
| Tier-only ranking | Top 10 are just the 10 highest-tier skills regardless of connectivity/breadth | All 4 scoring dimensions must contribute |
| No justification | Picks listed without explaining why they're useful | "Why top 10?" answer is mandatory for each |
| Stale results | Same 10 every time regardless of library changes | Score should reflect current skills.json state |
| Missing functions | All picks are analytical — no decision, planning, or creation skills | Function check is mandatory |
| Line count gaming | Long but shallow skill scores high on depth proxy | Depth proxy is line_count, but justification checks actual quality |
Depth Scaling
| Depth | Scoring | Ranking | Justification |
|---|
| 1x | Tier + connectivity only | Top 10 with category cap | 1-line descriptions |
| 2x | All 4 dimensions | Top 10 + diversity + function check | Full justifications + honorable mentions |
| 4x | 4 dimensions + quality audit of candidates | Optimized ranking with diversity/function iteration | Justifications + notable absences + comparison |
| 8x | Full library audit with quality-adjusted scores | Multi-pass ranking with cross-validation | Defensible essay per pick + alternative top-10 sets |
Default: 2x. These are floors.
Pre-Completion Checklist
Integration
- Shortcut for:
/pick 10 useful
- Use when: You want the objectively best general-purpose skills
- Routes to: The 10 picked skills;
/p10diverse for coverage-maximizing alternative
- Related:
/p10diverse (coverage over quality), /p10random (serendipity over quality)
- Differs from /p10diverse: diverse maximizes coverage; useful maximizes quality
- Differs from /p10random: random has no quality bias; useful is entirely quality-driven
- Differs from /p10goal: goal matches to a specific purpose; useful is purpose-agnostic
