feat(cli): add skill tools and refresh commands #454

Merged

aditya merged 2 commits from feature/m4-skill-registry-cli into master

2026-02-27 12:15:08 +00:00

aditya commented

2026-02-26 09:59:24 +00:00

Member

Summary

Implements CLI commands for skill refresh and enhances existing skill commands with capability summary information.

Changes

agents skill refresh <name>: Recomputes tool flattening for a single skill and syncs MCP-backed skills
agents skill refresh --all: Refreshes all registered skills in the registry
Enhanced skill list: Added Description column in rich output and capability_summary field in JSON/YAML formats
Enhanced skill show: Added Capability Summary panel in rich output with total tools, read-only status, and MCP server count
Enhanced skill tools: Added Summary panel displaying capability information alongside flattened tool list
Format Support: Added --format json/yaml output schemas for all commands with consistent capability_summary structure
Error Handling: CLI errors for nonexistent skills during refresh and helpful messages for empty registries
BDD Tests: Behave scenarios in features/skill_cli.feature covering refresh commands and enhanced outputs
Robot Tests: Integration test cases in robot/skill_cli.robot for CLI smoke tests
ASV Benchmarks: Performance benchmarks in benchmarks/skill_cli_bench.py for CLI overhead baseline (refresh operations: 14-25ms)
Documentation: Reference documentation at docs/reference/skill_cli.md with refresh command examples, output columns, and caching behavior

Motivation

The skill refresh command provides a mechanism to recompute tool flattening after skill modifications and sync MCP-backed skills without re-registering them. Enhanced outputs with capability summaries give users immediate visibility into skill capabilities (tool count, read-only status, MCP server presence) across all CLI commands.

Key Design Decisions

Refresh command performs Agent Skills discovery when skills.agent_skills_paths is configured, ensuring auto-discovery stays synchronized
Capability summary computation uses existing SkillService.compute_capability_summary() for consistency
MCP sync currently marks as refreshed; full MCP adapter integration deferred to MCP milestone
All enhanced outputs maintain backward compatibility with existing rich output while adding structured fields to JSON/YAML
Error messages follow CLI standards with helpful guidance (e.g., "Must specify either or --all")

Closes #167

## Summary Implements CLI commands for skill refresh and enhances existing skill commands with capability summary information. ## Changes - **`agents skill refresh <name>`**: Recomputes tool flattening for a single skill and syncs MCP-backed skills - **`agents skill refresh --all`**: Refreshes all registered skills in the registry - **Enhanced `skill list`**: Added Description column in rich output and capability_summary field in JSON/YAML formats - **Enhanced `skill show`**: Added Capability Summary panel in rich output with total tools, read-only status, and MCP server count - **Enhanced `skill tools`**: Added Summary panel displaying capability information alongside flattened tool list - **Format Support**: Added `--format json/yaml` output schemas for all commands with consistent capability_summary structure - **Error Handling**: CLI errors for nonexistent skills during refresh and helpful messages for empty registries - **BDD Tests**: Behave scenarios in `features/skill_cli.feature` covering refresh commands and enhanced outputs - **Robot Tests**: Integration test cases in `robot/skill_cli.robot` for CLI smoke tests - **ASV Benchmarks**: Performance benchmarks in `benchmarks/skill_cli_bench.py` for CLI overhead baseline (refresh operations: 14-25ms) - **Documentation**: Reference documentation at `docs/reference/skill_cli.md` with refresh command examples, output columns, and caching behavior ## Motivation The skill refresh command provides a mechanism to recompute tool flattening after skill modifications and sync MCP-backed skills without re-registering them. Enhanced outputs with capability summaries give users immediate visibility into skill capabilities (tool count, read-only status, MCP server presence) across all CLI commands. ## Key Design Decisions - Refresh command performs Agent Skills discovery when `skills.agent_skills_paths` is configured, ensuring auto-discovery stays synchronized - Capability summary computation uses existing `SkillService.compute_capability_summary()` for consistency - MCP sync currently marks as refreshed; full MCP adapter integration deferred to MCP milestone - All enhanced outputs maintain backward compatibility with existing rich output while adding structured fields to JSON/YAML - Error messages follow CLI standards with helpful guidance (e.g., "Must specify either <name> or --all") Closes #167

Rows
Columns