jbaruch/sub-agent-delegation
Meta-skill: how to pass skills/context to Claude sub-agents that start with fresh context, with documented SDK gotchas.
94
94%
Does it follow best practices?
Impact
96%
1.50xAverage score across 3 eval scenarios
Passed
No known issues
Quality
Discovery
100%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 is an excellent skill description that clearly articulates specific capabilities (sub-agent orchestration, skill inheritance, echo-skills handshake) across named platforms, provides explicit 'Use when' triggers covering multiple natural scenarios, and occupies a highly distinctive niche. The description is concise yet comprehensive, uses third-person voice correctly, and includes both positive use cases and troubleshooting scenarios.
| Dimension | Reasoning | Score |
|---|---|---|
Specificity | Lists multiple specific concrete actions: orchestrating sub-agents across two named platforms, covering inheritance behavior, passing skills explicitly, and the echo-skills validation handshake. These are concrete, actionable capabilities. | 3 / 3 |
Completeness | Clearly answers both 'what' (orchestrate sub-agents, covers inheritance, passing skills, echo-skills handshake) and 'when' with an explicit 'Use when' clause listing four specific trigger scenarios including spawning, delegating, building pipelines, and troubleshooting. | 3 / 3 |
Trigger Term Quality | Includes strong natural trigger terms users would say: 'sub-agents', 'child agents', 'delegating tasks', 'multi-agent pipelines', 'Claude Code', 'GitHub Copilot', 'skills', 'context'. Also covers troubleshooting scenarios which is a natural user need. | 3 / 3 |
Distinctiveness Conflict Risk | Highly distinctive niche: multi-agent orchestration specifically across Claude Code and GitHub Copilot with skill inheritance and validation handshake. Very unlikely to conflict with other skills due to the specific platform pairing and sub-agent focus. | 3 / 3 |
Total | 12 / 12 Passed |
Implementation
85%Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.
This is a strong, well-structured skill that addresses a genuinely complex cross-platform topic. It provides concrete, executable examples, a clear validation protocol with error recovery, and appropriate progressive disclosure to external references. The main weakness is minor verbosity in the Copilot section and some redundancy between the comparison table and inline descriptions.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The content is mostly efficient and covers genuinely non-obvious information (inheritance models, platform differences, validation protocol). However, some sections are slightly verbose — the VS Code Copilot section repeats the context-loss framing, and the cross-platform table partially duplicates inline content. A few sentences could be tightened. | 2 / 3 |
Actionability | Provides fully executable Python code for both AgentDefinition usage and the echo-skills validation handler. Concrete examples with real skill names, specific file paths for Copilot instructions, and exact #file reference syntax. The 'How to act' section gives a clear, numbered action plan. | 3 / 3 |
Workflow Clarity | The 'How to act' section provides a clear 4-step sequence. The validation protocol includes an explicit feedback loop — echo back skills, verify match, abort loudly on mismatch. The workflow covers the critical failure mode (context loss) with a concrete validation checkpoint before proceeding. | 3 / 3 |
Progressive Disclosure | The skill provides a well-structured overview with clear references to external files: `scripts/echo_skills.py` for the validation script, `docs/SUB_AGENT_ISSUES.md` for known issues, and `sub-agent-delegation-rules` for a quick reminder. References are one level deep and clearly signaled with relative paths. | 3 / 3 |
Total | 11 / 12 Passed |
Validation
100%Checks the skill against the spec for correct structure and formatting. All validation checks must pass before discovery and implementation can be scored.
Validation — 11 / 11 Passed
Validation for skill structure
No warnings or errors.
Reviewed
Table of Contents