agent-orchestration
Use when you need to launch and drive other AI agents (Claude Code, Aider, Codex, etc.) through their terminal interfaces via wsh. Examples: "run multiple Claude Code sessions in parallel on different tasks", "feed a task to an AI agent and handle its approval prompts", "coordinate several AI agents working on subtasks of a larger project".
68
81%
Does it follow best practices?
Impact
—
No eval scenarios have been run
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 a strong skill description that clearly defines a distinct niche—orchestrating AI agents through terminal interfaces. It provides concrete actions, natural trigger terms including specific tool names, and an explicit 'Use when' clause with illustrative examples. The description is concise yet comprehensive, making it easy for Claude to select this skill appropriately.
| Dimension | Reasoning | Score |
|---|---|---|
Specificity | Lists multiple specific concrete actions: launching AI agents, driving them through terminal interfaces via wsh, running parallel sessions, feeding tasks, handling approval prompts, and coordinating agents on subtasks. | 3 / 3 |
Completeness | Clearly answers both 'what' (launch and drive AI agents through terminal interfaces via wsh) and 'when' (explicit 'Use when' clause with concrete examples like running parallel sessions, feeding tasks, coordinating agents). | 3 / 3 |
Trigger Term Quality | Excellent coverage of natural terms users would say: 'Claude Code', 'Aider', 'Codex', 'AI agents', 'parallel', 'approval prompts', 'coordinate', 'subtasks'. These are terms users would naturally use when requesting multi-agent orchestration. | 3 / 3 |
Distinctiveness Conflict Risk | Highly distinctive niche focused on orchestrating other AI agents via wsh terminal interfaces. The specific mention of agent names (Claude Code, Aider, Codex), approval prompts, and parallel coordination makes it very unlikely to conflict with other skills. | 3 / 3 |
Total | 12 / 12 Passed |
Implementation
62%Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.
This is a well-structured orchestration guide with strong workflow clarity and practical patterns for managing AI agents via wsh sessions. Its main weaknesses are moderate verbosity (motivational content, explanations of obvious benefits) and pseudocode-style commands that aren't fully executable. The content would benefit from tightening prose and either providing fully executable examples or more clearly bridging to the actual tool/API interface.
Suggestions
Remove the 'Why?' section and motivational phrases ('This is not science fiction', 'You become a manager of AI workers') — Claude can infer the benefits from the patterns themselves.
Make command examples more concrete and executable — either show actual wsh_* tool calls or actual curl commands rather than descriptive pseudocode like 'create session "agent-auth" with command: ...'
Consider splitting the multi-agent coordination section and pitfalls into a separate ADVANCED.md or PATTERNS.md file, keeping SKILL.md as a concise overview with references.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The skill contains some unnecessary motivational framing ('This is not science fiction', 'You become a manager of AI workers') and the 'Why?' section explains benefits Claude can infer. However, the bulk of the content is practical patterns and guidance that adds genuine value. Could be tightened by ~20-30%. | 2 / 3 |
Actionability | Provides concrete pseudocode-style patterns (create session, send, read screen) and specific examples like auto-approve flags and git worktree commands. However, the commands are not fully executable — they use a descriptive/pseudocode style ('create session "agent-auth" with command: claude...') rather than actual tool calls or curl commands. The execution context header tries to bridge this gap but adds ambiguity. | 2 / 3 |
Workflow Clarity | The agent lifecycle is clearly sequenced (startup → thinking → working → approval → waiting). The delegation pattern provides an explicit 6-step workflow. Approval handling covers three strategies with clear decision points. Monitoring includes a polling loop pattern. Pitfalls section serves as validation/error-recovery guidance covering infinite loops, workspace conflicts, and cleanup. | 3 / 3 |
Progressive Disclosure | The content is well-structured with clear headers and logical sections, but it's a monolithic document (~200+ lines) with no references to supporting files. The execution context header references 'skills/core/SKILL.md' for API fallback, but the main body could benefit from splitting detailed patterns (multi-agent coordination, pitfalls) into separate reference files. No bundle files are provided to support progressive disclosure. | 2 / 3 |
Total | 9 / 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.
- Repository
- deepgram/wsh
- Commit
4863aaf
- Reviewed
Table of Contents
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.