History

Ollie Agent 2d74ffb95e prompt audit: trim system prompt, kill dead code, improve maintainability - System prompt: ~200 → ~80 lines (removed tool-first table, API docs, output protocol, dead 9P entries) - Removed AllowTools entirely (field, RPC, configs) - Removed PRIME_* env vars, replaced with typed struct fields - Merged tool listing + docs into single renderTools() - Killed BuildToolListing, changed OnToolsChanged to func() signal - Typed Preamble.Section (compile-time safety) - Added protocol docs on extractToolResult - Added session/ package godoc - Moved output protocol to per-agent prompts - Rewrote data/agents/README.md - Moved 7 reference docs from doc/resources/ to doc/ - Deleted PromptEnv() dead function		2026-08-09 14:13:52 +02:00
..
README.md	prompt audit: trim system prompt, kill dead code, improve maintainability	2026-08-09 14:13:52 +02:00
copilot.json	prompt audit: trim system prompt, kill dead code, improve maintainability	2026-08-09 14:13:52 +02:00
default.json	prompt audit: trim system prompt, kill dead code, improve maintainability	2026-08-09 14:13:52 +02:00
driver.json	prompt audit: trim system prompt, kill dead code, improve maintainability	2026-08-09 14:13:52 +02:00
explorer.json	remove hooks blocks and beads/freeloader refs from all agent JSON configs	2026-08-07 07:15:04 +02:00
librarian.json	prompt audit: trim system prompt, kill dead code, improve maintainability	2026-08-09 14:13:52 +02:00
navigator.json	remove hooks blocks and beads/freeloader refs from all agent JSON configs	2026-08-07 07:15:04 +02:00
taskmanager.json	prompt audit: trim system prompt, kill dead code, improve maintainability	2026-08-09 14:13:52 +02:00
theo.json	remove hooks blocks and beads/freeloader refs from all agent JSON configs	2026-08-07 07:15:04 +02:00

README.md

Agents

Each .json file defines an agent configuration. Agent configs are installed to ~/.config/ollie/agents/ and selected at runtime via /agent <name>.

Config format

{
  "prompt": ["$XDG_CONFIG_HOME/ollie/prompts/agent-coding.md"],
  "autoLoad": ["shell", "file_read", "file_edit", "file_glob", "file_grep"],
  "maxSteps": 25,
  "temperature": 0.5,
  "maxTokens": 16384,
  "backend": "openrouter",
  "model": "deepseek/deepseek-v4-flash",
  "systemPrompt": "/path/to/override.md",
  "compactionModel": "cheap-model-name"
}

Field	Purpose
`prompt`	File paths to concatenate as the agent prompt. Env vars expanded.
`autoLoad`	Tools loaded at session start.
`maxSteps`	Cap on tool-call rounds per turn. 0 = unlimited.
`temperature`	Sampling temperature.
`maxTokens`	Max output tokens per response.
`backend`	Override backend for this agent.
`model`	Override model for this agent.
`systemPrompt`	Override the embedded system prompt with a file path.
`compactionModel`	Use a cheaper model for context compaction.

Agents

Name	Role
`default`	Base coding agent. Also the base for JIT subagents.
`driver`	Full-featured agent with all tools, LSP, GUI, memory, subagents.
`copilot`	Read-only code copilot. Suggests changes as code fences.
`navigator`	Pair programming partner. Reads and advises, doesn't edit.
`explorer`	Code exploration and explanation. Read-only.
`librarian`	Document search and knowledge curation. Read-only.
`taskmanager`	TODO list management and planning.
`theo`	Security auditor (Theo de Raadt voice). Read-only.

System prompt assembly

The final system prompt sent to the model is assembled in sections:

system — embedded base prompt (prompts/system_prompt.md)
env — runtime environment (cwd, platform, date, git status)
agent — concatenated files from the prompt array
tools — auto-generated from loaded tools (name + .meta prompt)

The fully assembled prompt is readable at session/{sname}/agent/{aname}/systemprompt.

Adding a new agent

Copy default.json, rename it.
Adjust the prompt array to reference your prompt files in data/prompts/.
Set autoLoad to the tools this agent needs.
Set generation parameters as needed.
Run just install-data to install.