speculative-decoding

Accelerate LLM inference using speculative decoding, Medusa multiple heads, and lookahead decoding techniques. Use when optimizing inference speed (1.5-3.6× speedup), reducing latency for real-time applications, or deploying models with limited compute. Covers draft models, tree-based attention, Jacobi iteration, parallel token generation, and production deployment strategies.

Quality

75%

Does it follow best practices?

Run evals on this skill

Adds up to 20 points to the overall score

View guide

Securityby

Low

Low-risk findings worth noting

Fix and improve this skill with Tessl

tessl review fix ./backend/cli/skills/ml-inference/speculative-decoding/SKILL.md

Quality

Content

57%Weight 40%Scale 1-5

Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.

The body is actionable with solid code examples and a reasonable method-comparison structure, but it is padded with concepts Claude already knows, lacks validation checkpoints in its workflows, and has a broken reference path plus inlined detail that belongs in the bundled reference files.

Suggestions

Add the missing references/draft_model.md (or remove the broken See Also pointer) and weave reference links into the relevant Core Concepts sections instead of only listing them at the end.

Trim restated background Claude already knows (ASCII architecture trees, sequential-vs-Jacobi math reformulation) to tighten conciseness.

Add explicit validation/checkpoint steps in Best Practices (e.g., how to measure and confirm the claimed 1.5-3.6x speedup before committing to a method) to lift workflow clarity.

Dimension	Reasoning	Score
Conciseness	Mostly efficient code blocks, but padded with concepts Claude already knows (ASCII architecture trees, restating sequential-vs-Jacobi math, illustrative pseudocode), so it includes unnecessary explanation that could be tightened.	3 / 5
Actionability	Provides concrete, mostly executable code for transformers assisted generation, Medusa, lookahead, and vLLM, with minor gaps where examples are illustrative pseudocode rather than library-runnable.	4 / 5
Workflow Clarity	Presents a method-selection flow (Quick Start, comparison table, Best Practices if/elif) but lacks explicit validation checkpoints or feedback loops for verifying the claimed speedups in practice.	3 / 5
Progressive Disclosure	Has a See Also pointing to reference files, but one referenced path (references/draft_model.md) is missing from the bundle, references are only signaled at the end rather than at relevant sections, and substantial detail is inlined that overlaps the reference files.	3 / 5
	Total	13 / 20 Passed

Description

92%Weight 40%Scale 1-5

Based on the skill's description, can an agent find and select it at the right time? Clear, specific descriptions lead to better discovery.

The description is strong: it names concrete capabilities, provides an explicit 'Use when' trigger with multiple conditions, and occupies a distinct niche with low conflict risk. Trigger-term quality is the only dimension slightly below ceiling, leaning technical rather than covering lay synonyms.

Dimension	Reasoning	Score
Specificity	Lists multiple concrete capabilities across three named techniques plus a coverage clause ('draft models, tree-based attention, Jacobi iteration, parallel token generation, and production deployment strategies'), matching the comprehensive multi-action anchor.	5 / 5
Completeness	Explicitly answers both 'what' (accelerate LLM inference via speculative decoding, Medusa, lookahead) and 'when' via a concrete 'Use when...' clause with multiple trigger conditions.	5 / 5
Trigger Term Quality	Includes natural phrases a user would say ('optimizing inference speed', 'reducing latency for real-time applications', 'deploying models with limited compute') with good coverage, but leans technical and misses some common lay synonyms, falling just below the comprehensive anchor.	4 / 5
Distinctiveness Conflict Risk	Occupies a clear niche (speculative decoding / Medusa / lookahead decoding) with distinct, specific triggers and minimal overlap risk with other skills.	5 / 5
	Total	19 / 20 Passed

Validation

87%

Warnings & errors only

Checks the skill against the spec for correct structure and formatting. All validation checks must pass before discovery and implementation can be scored.

Validation — 14 / 16 Passed

Validation for skill structure

Criteria	Description	Result
frontmatter_unknown_keys	Unknown frontmatter key(s) found; consider removing or moving to metadata	Warning
referenced_paths_exist	Referenced path issues: 1 missing	Warning

	Total	14 / 16 Passed

Repository: synthetic-sciences/openscience
Path: backend/cli/skills/ml-inference/speculative-decoding/SKILL.md
Commit: fcd2e11

Reviewed: 1 day ago

Table of Contents

Discovery Implementation Validation

Is this your skill?

If you maintain this skill, you can claim it as your own. Once claimed, you can manage eval scenarios, bundle related skills, attach documentation or rules, and ensure cross-agent compatibility.