skill-council
Run a configurable multi-LLM council with personas, budget caps, synthesis, veto gates, and optional implementation handoff.
What this skill does
# Council
Use this skill for `/octo:council` and council-style requests.
## EXECUTION CONTRACT: MANDATORY
`/octo:council` is a runner-backed workflow. Execute the real shell runner by default:
```bash
"${CLAUDE_PLUGIN_ROOT:-$HOME/.claude-octopus/plugin}/scripts/orchestrate.sh" council <user arguments>
```
Do not simulate a council, role-play all members inside the current model, or answer directly in place of the runner. A single-model simulation must be explicitly requested with `--simulate` or `--single-model`; when used, label the output as `single-model simulation` and keep the generated `summary.json` path visible. Otherwise, missing provider quorum is a partial/failed council, not permission to silently downgrade.
## Phase 0A: Interactive Clarification
If the user invokes `/octo:council` with an ambiguous task, missing goal/depth intent, uncertain implementation permission, or unclear research/corpus expectations, use `AskUserQuestion` before running the council runner. If the prompt already includes clear flags and a clear task, skip this phase and execute immediately.
Use 2-4 mutually exclusive choices per question, put the recommended option first, and map the answers to runner flags before execution. Do not end with a free-form question or a set of questions; present the choice interaction and wait for the answer.
Default council clarification:
```javascript
AskUserQuestion({
questions: [
{
question: "How should the council handle this request?",
header: "Council Goal",
multiSelect: false,
options: [
{label: "Advice (Recommended)", description: "Return a structured recommendation without implementation"},
{label: "Decision", description: "Optimize for choosing between specific options"},
{label: "Implementation plan", description: "Produce a plan but do not edit files"},
{label: "Review", description: "Critique existing code, docs, or strategy"}
]
},
{
question: "How deep should the council go?",
header: "Depth",
multiSelect: false,
options: [
{label: "Standard (Recommended)", description: "Balanced cost and coverage"},
{label: "Quick", description: "Faster, lower-cost pass"},
{label: "Deep", description: "More critique and revision, higher cost"}
]
},
{
question: "Should the council use project research or corpus storage?",
header: "Context",
multiSelect: false,
options: [
{label: "No corpus (Recommended)", description: "Run with the provided prompt and write normal run artifacts"},
{label: "Research first", description: "Add --research-first before provider fanout"},
{label: "Append corpus", description: "Use --corpus-mode append for durable project notes"},
{label: "Require corpus", description: "Use --corpus-mode require and stop if no corpus workspace exists"}
]
}
]
})
```
Translate answers as follows:
- Advice, Decision, Implementation plan, or Review -> `--goal advice|decision|plan|review`
- Quick, Standard, or Deep -> `--depth quick|standard|deep`
- Research first -> `--research-first`
- Append corpus or Require corpus -> `--corpus-mode append|require`
## Phase 0: Preflight
Collect or infer:
- goal: `advice`, `decision`, `plan`, `implement`, or `review`
- domain: `auto`, `architecture`, `product`, `security`, `business`, `research`, or `docs`
- style: `balanced`, `adversarial`, `implementation`, `executive`, or `red-team`
- depth: `quick`, `standard`, or `deep`
- members: `auto`, `3`, `5`, or `7`
- budget cap in USD
- providers and provider availability
- pinned personas
- implementation permission and worktree isolation
- whether research must happen before council fanout (`--research-first`)
- whether findings, research, synthesis, and plans must be appended to the project corpus (`--corpus-mode append|require`)
Show the selected council, provider availability, benchmark freshness, quorum requirement, and cost estimate before provider fanout. Re-check the budget before critique, revision, synthesis, and implementation planning so the run stops before the next phase would exceed `--max-cost`. If the run is a dry run, stop after this preflight and write `summary.json`.
If `--research-first` is set, gather local corpus evidence first, then current external sources when allowed. Store concise findings as run artifacts before provider fanout. If `--corpus-mode append` or `--corpus-mode require` is set, append the research notes, findings, synthesis, and implementation plan to the project's durable corpus conventions before claiming completion; `require` must stop when no corpus workspace is available.
## Quorum
- quick requires at least one non-chair response plus a synthesis-capable chair
- standard and deep require at least two non-chair responses plus a synthesis-capable chair
- if the chair fails, retry once with the highest-scoring synthesis-capable fallback
- if quorum is lost, stop by default and present partial artifacts instead of pretending consensus exists
## Council Procedure
1. **Independent advice:** ask each selected persona for recommendation, assumptions, risks, implementation notes, and confidence.
2. **Cross-critique:** for standard/deep runs, semi-anonymize responses and ask members to critique gaps, assumptions, and risks.
3. **Revision:** for deep runs or high disagreement, let members revise their positions after critique.
4. **Chair synthesis:** produce agreement, disagreement, minority reports, risk register, implementation path, confidence, and conditions that would change the recommendation.
5. **Ratify / veto:** verifier, red-team, security, legal, finance, and medical roles can veto implementation for critical risks.
## Implementation Gates
- **Gate A:** user accepts the council synthesis or asks for revision.
- **Gate B:** user accepts the concrete implementation plan generated from the synthesis.
- **Gate C:** execution proceeds through existing Octopus implementation safety behavior. This is a one-shot authorization for the accepted plan, not per-file approval, unless existing safety hooks detect destructive or risky actions.
Never implement from council output without explicit approval. Preserve disagreement, summarize risks, and keep vetoes visible in the final answer and artifacts.
## Interactive Choice Handling
When the user needs to clarify scope, select a depth/provider/corpus option, approve a gate, or decide between follow-up actions, use `AskUserQuestion` with 2-4 mutually exclusive choices. Put the recommended choice first. Do not end with a free-form question or a set of questions; present the choice interaction and wait for the answer.
Related in AI Agents
skill-development
IncludedComprehensive meta-skill for creating, managing, validating, auditing, and distributing Claude Code skills and slash commands (unified in v2.1.3+). Provides skill templates, creation workflows, validation patterns, audit checklists, naming conventions, YAML frontmatter guidance, progressive disclosure examples, and best practices lookup. Use when creating new skills, validating existing skills, auditing skill quality, understanding skill architecture, needing skill templates, learning about YAML frontmatter requirements, progressive disclosure patterns, tool restrictions (allowed-tools), skill composition, skill naming conventions, troubleshooting skill activation issues, creating custom slash commands, configuring command frontmatter, using command arguments ($ARGUMENTS, $1, $2), bash execution in commands, file references in commands, command namespacing, plugin commands, MCP slash commands, Skill tool configuration, or deciding between skills vs slash commands. Delegates to docs-management skill for official documentation.
reprompter
IncludedTransform messy prompts into well-structured, effective prompts — single or multi-agent. Use when: "reprompt", "reprompt this", "clean up this prompt", "structure my prompt", rough text needing XML tags and best practices, "reprompter teams", "repromptception", "run with quality", "smart run", "smart agents", multi-agent tasks, audits, parallel work, anything going to agent teams. Don't use when: simple Q&A, pure chat, immediate execution-only tasks. See "Don't Use When" section for details. Outputs: Structured XML/Markdown prompt, quality score (before/after), optional team brief + per-agent sub-prompts, agent team output files. Success criteria: Single mode quality score ≥ 7/10; Repromptception per-agent prompt quality score 8+/10; all required sections present, actionable and specific.
adaptive-compaction
IncludedAdaptive add-on policy and recovery layer that decides WHEN to compact, prune, snapshot, or fork -- replacing fixed-percent auto-compaction across Claude Code, Codex, and MCP-capable hosts. Trigger on auto-compact timing or damage: "when should I compact", "is it safe to compact now or start a fresh session", "auto-compact fires too early/mid-task", "switching to an unrelated task but the window still has space", "context rot", "answers get worse the longer the session runs", "the agent forgot the plan or my decisions after it summarized", "add a layer on top that manages context without changing the agent", raising autoCompactWindow to give the policy room, or installing/tuning a cross-tool compaction policy or PreCompact hook -- even when "compaction" is never said but the problem is context-window pressure or post-summarization memory loss. Do NOT use to summarize a conversation, build RAG, write a summarization prompt (decides WHEN not HOW), or answer max-context-length trivia.
agent-skill-creator
IncludedCreate cross-platform agent skills from workflow descriptions. Activates when users ask to create an agent, automate a repetitive workflow, create a custom skill, or need advanced agent creation. Triggers on phrases like create agent for, automate workflow, create skill for, every day I have to, daily I need to, turn process into agent, need to automate, create a cross-platform skill, validate this skill, export this skill, migrate this skill. Supports single skills, multi-agent suites, transcript processing, template-based creation, interactive configuration, cross-platform export, and spec validation.
llm-wiki
IncludedUse when building or maintaining a persistent personal knowledge base (second brain) in Obsidian where an LLM incrementally ingests sources, updates entity/concept pages, maintains cross-references, and keeps a synthesis current. Triggers include "second brain", "Obsidian wiki", "personal knowledge management", "ingest this paper/article/book", "build a research wiki", "compound knowledge", "Memex", or whenever the user wants knowledge to accumulate across sessions instead of being re-derived by RAG on every query.
skill-master
IncludedAgent Skills authoring, evaluation, and optimization. Create, edit, validate, benchmark, and improve skills following the agentskills.io specification. Use when designing SKILL.md files, structuring skill folders (references, scripts, assets), ingesting external documentation into skills, running trigger evals, benchmarking skill quality, optimizing descriptions, or performing blind A/B comparisons. Keywords: agentskills.io, SKILL.md, skill authoring, eval, benchmark, trigger optimization.