launch-sub-agent

Launch an intelligent sub-agent with automatic model selection based on task complexity, specialized agent matching, Zero-shot CoT reasoning, and mandatory self-critique verification

Quality

28%

Does it follow best practices?

Impact

—

No eval scenarios have been run

Securityby

Passed

No known issues

Fix and improve this skill with Tessl

tessl review fix ./plugins/sadd/skills/launch-sub-agent/SKILL.md

Quality

Content

39%

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

The skill has excellent workflow structure with a clear 5-phase process and good decision logic for model selection, but it is severely bloated. The CoT reasoning template, self-critique template, and verbose explanations of concepts Claude already understands (like what Chain-of-Thought reasoning is) consume excessive tokens. The content would benefit enormously from being split across multiple files and trimmed of redundant explanations.

Suggestions

Reduce the CoT prefix template to 3-5 lines of instruction rather than a full markdown template — Claude knows how to reason step-by-step without a detailed scaffold.

Condense the self-critique suffix to a concise checklist (e.g., '1. Generate 5 verification questions specific to task. 2. Answer each with evidence. 3. Fix any gaps before submitting.') instead of the ~50-line template.

Extract the prompt templates (CoT prefix, task body, critique suffix) into separate bundle files and reference them from SKILL.md to improve progressive disclosure.

Remove the duplicate model selection presentation — keep either the table or the ASCII decision tree, not both.

Dimension	Reasoning	Score
Conciseness	Extremely verbose at ~250+ lines. The CoT prefix, self-critique suffix, and task body templates are massive boilerplate that Claude doesn't need spelled out in such detail. The decision tree is presented twice (table + ASCII art). Concepts like 'context isolation' and 'Zero-shot Chain-of-Thought' are explained rather than just applied. The self-critique section alone is ~50 lines of template that could be condensed to 5-10 lines of instruction.	1 / 3
Actionability	The skill provides structured guidance with decision trees and examples, but the actual dispatch mechanism is vague ('Use Task tool') without showing a concrete, executable invocation. The prompt templates are detailed but are more like fill-in-the-blank scaffolds than copy-paste ready commands. The specialized agent integration says 'read the agent definition' without showing how.	2 / 3
Workflow Clarity	The 5-phase workflow is clearly sequenced (Analyze → Select Model → Match Agent → Construct Prompt → Dispatch) with explicit decision points at each phase. The self-critique loop provides a clear validation/feedback mechanism with fix-and-re-verify steps. The decision tree provides unambiguous branching logic for model selection.	3 / 3
Progressive Disclosure	No bundle files are provided, yet the skill is a monolithic wall of text exceeding 250 lines. The massive CoT prefix template, self-critique suffix template, and detailed examples could all be split into referenced files. Everything is inlined, making the skill overwhelming to parse in a single read.	1 / 3
	Total	7 / 12 Passed

Description

17%

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

This description reads like a feature list of internal architecture rather than a user-facing skill description. It lacks natural trigger terms users would actually say, provides no guidance on when Claude should select this skill, and uses heavy technical jargon (Zero-shot CoT, self-critique verification) that obscures rather than clarifies its purpose.

Suggestions

Add an explicit 'Use when...' clause describing the scenarios that should trigger this skill, e.g., 'Use when the user asks to delegate a complex task, run a sub-task, or needs multi-step reasoning with verification.'

Replace technical jargon with natural language trigger terms users would actually say, such as 'delegate', 'run a subtask', 'break this down', 'think step by step', or 'verify your answer'.

Clarify the concrete outcome or value — what does the user get? E.g., 'Delegates complex tasks to a specialized sub-agent that reasons step-by-step and verifies its own output before returning results.'

Dimension	Reasoning	Score
Specificity	It names several concepts (sub-agent launching, model selection, agent matching, CoT reasoning, self-critique) but these are abstract architectural features rather than concrete user-facing actions. It describes how it works internally rather than what it produces.	2 / 3
Completeness	It partially addresses 'what' (launch a sub-agent) but provides no 'when' guidance whatsoever — there is no 'Use when...' clause or equivalent trigger guidance, and the description doesn't clarify what types of user requests should invoke this skill.	1 / 3
Trigger Term Quality	The terms used ('Zero-shot CoT reasoning', 'mandatory self-critique verification', 'specialized agent matching') are technical jargon that users would almost never naturally say. Users would more likely say 'run a task', 'delegate', or 'think harder about this'.	1 / 3
Distinctiveness Conflict Risk	The concept of launching a sub-agent with model selection is somewhat distinctive, but 'task complexity' and 'reasoning' are broad enough to potentially overlap with many other skills that involve delegation, planning, or multi-step problem solving.	2 / 3
	Total	6 / 12 Passed

Validation

90%

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 — 10 / 11 Passed

Validation for skill structure

Criteria	Description	Result
frontmatter_unknown_keys	Unknown frontmatter key(s) found; consider removing or moving to metadata	Warning

	Total	10 / 11 Passed

Repository: NeoLabHQ/context-engineering-kit
Commit: 3711edf

Reviewed: 20 days 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.