Goal
Add downloadable, service-specific skill/agent bundles for each configured MCP service entry.
Motivation
Agents should not have to poke around to discover commands, schemas, auth quirks, or safe usage patterns. A service entry should be able to publish a compact skill bundle that an agent downloads once, loads, and immediately knows how to call that service through mcp2cli.
Proposed shape
For each service in the registry, expose/export a bundle containing:
- service description and endpoint shape
- available tools and exact mcp2cli commands
- schemas summarized into agent-friendly examples
- auth/config requirements without secrets
- safe read/write boundaries
- common workflows
- known failure modes and error meanings
- freshness/version/hash metadata
Potential commands:
mcp2cli skills list
mcp2cli skills get open-brain --format hermes-skill > open-brain-mcp/SKILL.md
mcp2cli skills install open-brain --target ~/.hermes/skills/mcp/open-brain
mcp2cli skills diff open-brain
Acceptance criteria
- Skill output is deterministic and safe to commit/share when no secrets are present.
- Skill output includes exact runnable mcp2cli commands for each tool.
- Generated skills include schema/version/hash so agents can detect drift.
- Generated skills can be installed into Hermes skill directories or downloaded as an archive.
- Bundles can include service-specific examples like Open Brain session lanes, n8n workflow ops, Fleet Board task updates, etc.
Goal
Add downloadable, service-specific skill/agent bundles for each configured MCP service entry.
Motivation
Agents should not have to poke around to discover commands, schemas, auth quirks, or safe usage patterns. A service entry should be able to publish a compact skill bundle that an agent downloads once, loads, and immediately knows how to call that service through mcp2cli.
Proposed shape
For each service in the registry, expose/export a bundle containing:
Potential commands:
Acceptance criteria