party-mode
Installation
SKILL.md
Party Mode
Run a round-table where these personas talk to each other and to the user like real, distinct people in conversation. You're the orchestrator.
Conventions
- Paths: bare paths (e.g.
references/create-party.md) resolve from{skill-root}(the folder this SKILL.md lives in);{skill-root}/...means the skill folder itself. Any{skill-root}placeholder insideparty.toml(inoutput_dir) is resolved to the real absolute path byresolve_party.py— it returns absolute values. - Config: the skill's configuration and roster live in
party.tomlat the skill root. An optional user override atuser/party.user.tomlmerges over it (override wins). Custom members and groups you author for the user are written there too — it's the skill's own file, no external tooling. - Scripts (run via plain
python3, stdlib only):scripts/resolve_party.pyresolves the roster,party_mode,memory_enabled, and scene/open_cast. Memory is managed directly via Write/Edit tools — write session memory files todocs/party-mode-memories/and maintainindex.md. - File roles: a party's memory lives in per-session timestamped logs at
docs/party-mode-memories/with anindex.mdfor quick lookup; custom members and groups live inuser/party.user.toml. Mechanics inreferences/party-memory.md(memory),references/create-party.md(authoring), andreferences/model-selection.md(model assignment for subagent modes). - Search: Web-search, don't guess — anything past your cutoff or unfamiliar; subagents too.
- Output directories (fixed, relative to workspace root):
- Discussion memory:
docs/party-mode-memories/— unified memory directory. Containsindex.md(session index) and per-session memory files ({YYYY-MM-DD}-{topic}-memory.md). - Meeting reports & visual reviews:
docs/party-mode-report/— stores the meeting minutes report (.md) and the visual discussion review (.html). - These directories are not configurable via
party.toml; they are always relative to the current workspace root. Create them on first use if they don't exist.
- Discussion memory: