# Agent skills

Skills are public Markdown. Point an assistant at the URL instead of re-explaining DoorOps every session.

| Skill | Audience | URL |
| --- | --- | --- |
| `brief-doorops` | Prospects and evaluators | [doorops.com/SKILL.md](https://doorops.com/SKILL.md) |
| `operate-doorops` | Authenticated operators | [doorops.com/skills/operate-doorops.md](https://doorops.com/skills/operate-doorops.md) |

`brief-doorops` uses verified product facts only. It prepares a consent-bound enquiry. It never invents prices, customers, or delivery claims.

`operate-doorops` connects to tenant MCP with OAuth, calls `whoami` first, then reads or creates within role limits. AI summaries are review aids, not certificates.

## Discovery

- `/agent-skills/index.json`
- `/.well-known/skills/index.json`
- `/llms.txt`
- `/.well-known/agent.json`

Hashes in the skill index exist so a client can tell if a skill file changed.