CtrlK
BlogDocsLog inGet started
Tessl Logo
SkillPublic

multi-ai-research

Comprehensive research and analysis using Claude (subagents), Gemini CLI, and Codex CLI. Multi-perspective research with cross-verification, iterative refinement, and 100% citation coverage. Use for security analysis, architecture research, code quality assessment, performance analysis, or any research requiring rigorous verification and multiple AI perspectives.

Quality

64%

Does it follow best practices?

Run evals on this skill

Adds up to 20 points to the overall score

View guide
SecuritybySnyk

Low

Low-risk findings worth noting

Fix and improve this skill with Tessl

tessl review fix ./.claude/skills/multi-ai-research/SKILL.md

The canonical home for this skill is multi-ai-research in fernandezbaptiste/Skrillz

SKILL.md
Quality
Evals
Security

Quality

Content

48%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 pipeline architecture is genuinely well designed — clearly sequenced phases with explicit quality gates, verification, and iteration loops — but the document undermines itself with severe redundancy, marketing padding, and a missing layer of executable detail for actually dispatching work to the three AI systems. It reads as a promotional spec for a process rather than lean operational instructions.

Suggestions

Merge the duplicated 'When to Use' and 'Analysis Types' sections (same five categories listed twice) and cut marketing material ('Key Innovation', 'Quality Guarantees', fabricated 'Quality: 97/100' walkthroughs) — this alone could halve the file.

Add the concrete Phase 2 dispatch commands/prompts for Gemini CLI, Codex CLI, and Task-tool subagents so the pipeline is executable, not just described.

Move the three worked examples, CLI setup instructions, and troubleshooting into references/ files (e.g. references/examples.md, references/setup.md), keeping SKILL.md as a lean overview.

DimensionReasoningScore

Conciseness

A 730-line body with pervasive redundancy: 'When to Use' and 'Analysis Types' enumerate the same five categories twice with near-identical bullets, plus marketing sections ('Key Innovation', 'Quality Guarantees'), three fabricated result walkthroughs ('Quality: 97/100'), and unquarantined time-sensitive details ('latest 2024-2025', 'gemini-2.5-flash', 'o3'). Padding is severe, though it is redundancy rather than explaining concepts Claude already knows.

2 / 5

Actionability

Setup and invocation are executable (npm installs, `gemini`/`codex` auth flows, the verified `bash .claude/skills/multi-ai-research/scripts/analyze.sh "objective"` launcher), but the core pipeline is described rather than instructed: Phase 2 never provides the actual commands or prompts to dispatch research to Gemini/Codex or to spawn the named subagents. The key execution details are missing.

3 / 5

Workflow Clarity

Six labeled phases each carry a duration, an output artifact under `.analysis/`, explicit validation (3-pass verification, quality gates at 95), and a feedback loop (iterate until thresholds are met, with a troubleshooting section mapping failures to fixes). Docked one point because the header says 'The 5-Phase Pipeline' while listing six phases, and individual phase instructions remain abstract.

4 / 5

Progressive Disclosure

Section headers make the 730-line document navigable and the single bundle file (`scripts/analyze.sh`) is real and correctly referenced, but roughly 400 lines of clearly externalizable content (three worked examples, duplicated analysis-type catalogs, CLI setup, troubleshooting, quick-reference tables) are inlined with no references/ split.

3 / 5

Total

12

/

20

Passed

Description

66%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.

The description answers both what and when with genuinely natural trigger phrases, but is diluted by over-claim language ('Comprehensive', '100% citation coverage', 'rigorous') and a broad catch-all trigger that raises conflict risk. Tightening the 'what' to concrete deliverables and narrowing the trigger scope would lift it from good to excellent.

Suggestions

Replace over-claims ('Comprehensive', '100% citation coverage', 'rigorous verification') with concrete deliverables, e.g. 'Runs a multi-phase analysis pipeline and produces a cited final report with quality scoring'.

Narrow the catch-all 'or any research requiring rigorous verification and multiple AI perspectives' to specific triggers, reducing overlap with generic research/analysis skills.

Add natural trigger synonyms users would say, such as 'security audit', 'code review', 'vulnerability assessment', 'architecture review'.

DimensionReasoningScore

Specificity

Names the concrete tooling ('Claude (subagents), Gemini CLI, and Codex CLI') and a couple of actions ('cross-verification, iterative refinement'), but the action language leans on marketing abstractions ('Comprehensive research and analysis', '100% citation coverage') rather than listing several specific concrete actions. Over-claim phrasing keeps it below the several-specific-actions anchor.

3 / 5

Completeness

Both a 'what' (multi-AI research and analysis with cross-verification via three named AI systems) and an explicit 'Use for...' clause with five concrete trigger scenarios are present. The 'what' is somewhat vague ('Comprehensive research and analysis'), so it does not clearly meet the anchor-5 bar.

4 / 5

Trigger Term Quality

'security analysis, architecture research, code quality assessment, performance analysis' are natural phrases users would actually say. Common synonyms like 'audit', 'code review', or 'vulnerability assessment' are missing, so it falls short of comprehensive synonym coverage.

4 / 5

Distinctiveness Conflict Risk

The three-AI-systems framing is a distinctive niche, but the catch-all trigger 'or any research requiring rigorous verification and multiple AI perspectives' creates real overlap risk with generic research/analysis skills.

3 / 5

Total

14

/

20

Passed

Validation

87%

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

Validation — 14 / 16 Passed

Validation for skill structure

CriteriaDescriptionResult

skill_md_line_count

SKILL.md is long (736 lines); consider splitting into references/ and linking

Warning

allowed_tools_field

'allowed-tools' contains unusual tool name(s)

Warning

Total

14

/

16

Passed

Repository

fernandezbaptiste/Skrillz

Path

.claude/skills/multi-ai-research/SKILL.md

Revision

93ed392

Reviewed

Table of Contents

DiscoveryImplementationValidation

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.

Claim skill