Claude-skill-registry agent-evaluator

Deterministic custom subagent selection helper. Use when you need a reproducible, auditable decision on which custom subagents to activate for a user query (runs scripts/agent_evaluator.py).

install

source · Clone the upstream repo

git clone https://github.com/majiayu000/claude-skill-registry

Claude Code · Install into ~/.claude/skills/

T=$(mktemp -d) && git clone --depth=1 https://github.com/majiayu000/claude-skill-registry "$T" && mkdir -p ~/.claude/skills && cp -r "$T/skills/data/agent-evaluator" ~/.claude/skills/majiayu000-claude-skill-registry-agent-evaluator && rm -rf "$T"

manifest: skills/data/agent-evaluator/SKILL.md

Agent Evaluator

Evaluate a user query against the workspace's available subagents and return a JSON decision payload (activated/required/suggested agents and scoring).

Mechanism

Run the evaluator script (located in scripts folder relative this skill file) with the user query as an argument.:

python scripts/agent_evaluator.py "YOUR_QUERY_HERE"

Optional: include a contextual file path as the second argument:

python scripts/agent_evaluator.py "YOUR_QUERY_HERE" "path/to/file.ext"

Output

Writes a JSON object to stdout.

Key fields include:

```
activated_agents
```
```
required_agents
```
```
suggested_agents
```
```
evaluations
```
(per-agent score + reasoning)

Examples

Evaluate a query:

python scripts/agent_evaluator.py "Please help refine our custom instruction file"

Evaluate a query with file context:

python scripts/agent_evaluator.py "Update this instruction" "instructions/agent-forced-eval.instructions.md"