llm-council

Provider-agnostic multi-LLM deliberation. Three phases — independent responses, cross-model anonymized ranking, chairman synthesis. Provider config from env (OPENAI/ANTHROPIC/FIREWORKS/OPENROUTER/custom OpenAI-compatible base URL). Persists transcript to a wiki page when --wiki <slug> is passed. Use when the user wants multiple AI perspectives, consensus-building, or the "LLM Council" approach for high-stakes reviews, plan critique, or contested learning rules.

Quality

78%

Does it follow best practices?

Run evals on this skill

Adds up to 20 points to the overall score

View guide

Securityby

Passed

No findings from the security scan

Fix and improve this skill with Tessl

tessl review fix ./skills/llm-council/SKILL.md

Quality

Content

68%Weight 40%Scale 1-5

Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.

The body is a well-structured, mostly executable overview backed by a real bundle script, with lean formatting and clear phase sequencing. Its main weakness is the absence of validation/cost-confirm checkpoints in a batch, cost-incurring workflow, which caps workflow clarity.

Suggestions

Add an explicit pre-flight validation checkpoint before running — e.g., confirm required API keys are set and surface estimated call count/cost (len(models)^2 + chairman) for user confirmation, especially when more than 5 models are requested.

Include a brief failure/feedback loop: check that all phase-1 responses succeeded before proceeding to ranking, and re-run or skip a model if its response errors out.

Move the default per-provider model roster (or a representative example like '--models gpt-4o,claude-sonnet-5,deepseek-v3') inline or into a clearly signaled reference so the run command is fully copy-paste ready without opening council.js.

Dimension	Reasoning	Score
Conciseness	The body is lean and uses compact tables, command blocks, and a directory tree rather than prose, with only minor padding such as the 'dair-academy's version hardcoded Fireworks' aside, so it is efficient with trimmable over-explanation rather than verbose.	4 / 5
Actionability	Provides concrete executable commands ('node $SKILL_ROOT/scripts/council.js run "<query>" [--models ...] [--chairman ...] [--provider ...] [--wiki ...]'), a providers table, and an output tree pointing to the real council.js script, with minor gaps where the model roster is deferred to the script.	4 / 5
Workflow Clarity	The three phases (Independent, Ranking, Synthesis) are clearly sequenced and reinforced by 'Hard rules', but there are no validation checkpoints for what is inherently a batch, cost-incurring operation (N-squared ranking calls), and the rubric caps batch operations without validation/feedback loops at 3.	3 / 5
Progressive Disclosure	Well-organized into clearly headed sections with a one-level-deep, well-signaled reference to the verified scripts/council.js for default model rosters; navigation is easy, though the inlined env table and output tree represent minor organization gaps rather than the clean reference-splitting of the level-5 anchor.	4 / 5
	Total	15 / 20 Passed

Description

87%Weight 40%Scale 1-5

Based on the skill's description, can an agent find and select it at the right time? Clear, specific descriptions lead to better discovery.

A strong, third-person description that clearly states capabilities and gives explicit 'Use when' triggers covering several natural phrases. It is comprehensive on completeness and distinctiveness, with only minor specificity and trigger-synonym gaps keeping it just below ceiling on those dimensions.

Dimension	Reasoning	Score
Specificity	Names several concrete actions — 'independent responses, cross-model anonymized ranking, chairman synthesis', env-based provider config, and wiki transcript persistence — but phrases them somewhat architecturally rather than as a comprehensive action list, so it sits below the level-5 anchor.	4 / 5
Completeness	It explicitly answers both what (three-phase deliberation, provider config, wiki persistence) and when ('Use when the user wants multiple AI perspectives, consensus-building, or the "LLM Council" approach for high-stakes reviews, plan critique, or contested learning rules') with concrete trigger phrases.	5 / 5
Trigger Term Quality	'multiple AI perspectives, consensus-building, LLM Council, high-stakes reviews, plan critique' are natural phrases a user might say, giving good keyword coverage, though a few common synonyms are missing and 'contested learning rules' leans jargon-heavy.	4 / 5
Distinctiveness Conflict Risk	'LLM Council' with a defined three-phase structure and provider-agnostic framing carves a clear niche with distinct triggers and minimal overlap with other skills.	5 / 5
	Total	18 / 20 Passed

Validation

93%

Warnings & errors only

Checks the skill against the spec for correct structure and formatting. All validation checks must pass before discovery and implementation can be scored.

Validation — 15 / 16 Passed

Validation for skill structure

Criteria	Description	Result
allowed_tools_field	'allowed-tools' contains unusual tool name(s)	Warning

	Total	15 / 16 Passed

Repository: rohitg00/pro-workflow
Path: skills/llm-council/SKILL.md
Commit: 7f7209d

Reviewed: 3 days ago

Table of Contents

Discovery Implementation Validation

Is this your skill?

If you maintain this skill, you can claim it as your own. Once claimed, you can manage eval scenarios, bundle related skills, attach documentation or rules, and ensure cross-agent compatibility.