skill distiller

Other Community

Skill notes

Reference Files

File · Contents · Load When references/distillation-patterns.md · Pattern catalog for converting reasoning to rules · Always

Prerequisites

The source skill must exist and pass package-evaluator at >= 70% Access to both the source model (Opus) and target model (Haiku/Sonnet) for validation The surrogate-verifier skill for cross-model assertion checking

Phase 1: Complexity Analysis

Score each section of the source SKILL.md for reasoning difficulty: Complexity Signal · Score · Distillation Action Decision tree with 3+ branches · HIGH · Convert to explicit if/then lookup table "Use judgment" or "consider context" · HIGH · Replace with concrete heuristic rules Multi-step inference chain · HIGH · Break into numbered atomic steps

Phase 2: Trace Collection

Execute the source skill with Opus on 5 representative tasks: Select tasks from evals/cases.yaml (positive cases) or generate new ones For each task, capture the full execution trace: Tool calls made (which tools, in what order) Intermediate reasoning visible in output

Phase 3: Pattern Extraction

From the collected traces, extract deterministic patterns: Decision paths — For each HIGH-complexity section, find the actual decisions Opus made across the 5 tasks. If Opus chose the same path in 4/5 cases, that path becomes the default rule Lookup tables — Where Opus applied domain knowledge, build explicit lookup tables (e.g., "if input contains SQL, use these patterns; if input contains Python, use those")

Fluxly profile Runtime and permissions are inferred by Fluxly static checks, not author declarations or runtime test results.
Domain
Other
Author-declared agents
No explicit declaration found; this is not inferred or tested compatibility
Static check
88 / 100 · heuristic scan, not runtime safety proof
Runtime evidence
Not run; no isolated runtime test evidence is currently recorded
Author / version / license
@Mathews-Tom · no license declared
Token usage
Lean
Setup complexity
Guided setup
External API key
Not required
Operating systems
Unspecified (assume cross-platform)
Runtime requirements
Python
Detected file/system behavior
  • Read-only
  • Write / modify
  • Shell exec
Network behavior
Local-only
Install commands
None (reference only)

Heads up: Tool permissions are not declared; actual access depends on the runtime.

01
Phase 1: Complexity Analysis

Score each section of the source SKILL.md for reasoning difficulty: Complexity Signal · Score · Distillation Action Decision tree with 3+ branches · HIGH · Convert to explicit if/then lookup table "Use judgment" or "consider context" · HIGH · Replace with concrete heuristic rules Multi-step inference chain · HIGH · Break into numbered atomic steps

02
Phase 2: Trace Collection

Execute the source skill with Opus on 5 representative tasks: Select tasks from `evals/cases.yaml` (positive cases) or generate new ones For each task, capture the full execution trace: Tool calls made (which tools, in what order) Intermediate reasoning visible in output

03
Phase 3: Pattern Extraction

From the collected traces, extract deterministic patterns: Decision paths — For each HIGH-complexity section, find the actual decisions Opus made across the 5 tasks. If Opus chose the same path in 4/5 cases, that path becomes the default rule Lookup tables — Where Opus applied domain knowledge, build explicit lookup tables (e.g., "if input contains SQL, use these patterns; if input contains Python, use those")

04
Phase 4: Distilled Rewrite

Rewrite the SKILL.md applying all distillation actions from Phase 1: Source Pattern · Distilled Replacement "Analyze the code and determine..." · "Check for these 5 specific patterns: [list]" "Use appropriate formatting" · "Output as a markdown table with columns: [A, B, C]" "Consider the context to decide..." · "If [condition A]: do X. If [condition B]: do Y. Default: Z"

05
Phase 5: Target Model Validation

Run the distilled skill on the target model (Haiku or Sonnet): Execute the same 5 tasks from Phase 2 with the distilled skill loaded Use the `surrogate-verifier` to generate assertions for each task output Compare pass rates: Metric · Source (Opus + original) · Target (Haiku + distilled) · Delta

06
Phase 6: Cross-Model Report

Produce the final comparison:

Discussion

Powered by GitHub Discussions. Sign in with GitHub to comment, react, or subscribe.