agentry/drafter
Drafts a QuickAgent from a description, searching the registry and the web; AI Review reads every result first.
Built into Agentry. The application runs this package itself; it is published here so you can read how it works. Its README says what Agentry gives it beyond its manifest.
Declared capabilities
Ships knowledge: drafting-guide (211.3 KB), in the release archive under knowledge/<name>/, scanned with the version.
Capabilities are the author's declaration and the basis for policy and scanning. The registry cannot enforce them at run time. Your runtime should.
MCP servers & tools
agentry run serves exactly the tools this contract declares, through a per-run MCP gateway. An undeclared capability is a tool that does not exist at run time.
Tools the runner grants
| Manifest declaration | Tool(s) granted |
|---|---|
| (always) | write_report — reports → the run's output folder |
| (always) | list_inputs, read_input, search_inputs — operator-provided input files, read-only |
MCP servers involved
- QuickAgent Tools — the runner's companion server, served through a per-run Docker MCP Gateway; its tool surface is exactly the table above.
In an embedded install
Installed into an existing coding agent (agentry install), the same contract maps to the host's built-in tools — enforcement there is the host's. Allowed tools: none declared (host defaults apply).
Install
agentry install agentry/drafter@1.15.4git clone https://secureagentry.net/agentry/drafter.gitgit clone ssh://git@git.secureagentry.net/agentry/drafter.gitclaude-skill is this agent's default target, so the flag is not needed.
knowledge/drafting-guide/wiki/manifest/run-modes.md
---
title: Run modes
type: guide
summary: Autonomous and interactive runs, when a draft adds interactive, what a conversation can and cannot do, and the opening line.
sources: [web:https://docs.secureagentry.net/quickagents/, web:https://docs.secureagentry.net/agentry/hosted-runs/]
updated: 2026-10-12
---
# Run modes
`modes` lists how a QuickAgent can run [src: web:https://docs.secureagentry.net/quickagents/]:
- `["autonomous"]`, the default: a run starts from the task and the person's files, after the intake's questions, does the job, returns its result and ends.
- `["interactive"]`: the run is a conversation. In the desktop it is the run panel's chat, in Agentry a session (*Talk to it*), in the CLI a chat in the terminal. The agent can ask for what the work turns out to need, and the person can answer and add files to the run's read-only inputs before asking for the final result.
- Both: the person chooses when they start it.
## When to add interactive
Add `interactive` only when the description asks for a conversation, a chat or back-and-forth: practising an interview, a tutor, a helper the person talks a problem through with. What the person names or gives the QuickAgent ("a city I name", "the notes I paste", "the repository I choose") is the task or an input of an autonomous run, not a reason for a conversation. A description that asks to approve each change before it is made (on a server, say) asks for one.
## What a conversation cannot do
A conversation has exactly the contract an autonomous run of the same release has: the same network, commands, keys, servers and files. Nothing the person says in it gives the agent more access, so declare in the manifest everything the conversation will use [src: web:https://docs.secureagentry.net/quickagents/].
## The opening line
With `interactive`, add `interactive.opening`: the first line the agent says when a session starts with no message from the person, at most 500 characters. Make it ask for what the work starts from ("Which incident, and what do you need from it?").
The instructions of an interactive QuickAgent say how to conduct the conversation in `## Steps`: what to ask first, when to ask again, and when the work is done and the final result is written [src: web:https://docs.secureagentry.net/agentry/hosted-runs/].