CtrlK
BlogDocsLog inGet started
Tessl Logo
SkillPublic

three-way-judge

Run an explicitly requested three-option workflow with three independent executors, one independent paired critic per option, bounded revision loops, and a blind evidence-backed judge. Use only when the user invokes $three-way-judge or explicitly requests this exact panel; this v1 skill does not run the nine-review deep mode.

Quality

92%

Does it follow best practices?

Run evals on this skill

Adds up to 20 points to the overall score

View guide
SecuritybySnyk

Passed

No findings from the security scan

SKILL.md
Quality
Evals
Security

Quality

Content

93%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.

An exceptionally well-engineered orchestration skill: lean imperative rules, executable validation commands at every stage, and a well-signaled one-level reference structure that fully exists on disk. The only flaw is a minor chronological misplacement of the panel-decision validation step within the command sequence.

Suggestions

Reorder the verification command block so 'node <skill-directory>/scripts/build-blind-packet.mjs' and the blind-judging step appear before 'node <skill-directory>/scripts/validate-panel.mjs', matching the actual pipeline chronology.

Consider adding a one-line explicit statement of the pipeline order (freeze, dispatch, critique, gate, blind packet, judge, reconcile, verify, report) at the top of the Runtime behavior section so the three sections assemble without reader inference.

DimensionReasoningScore

Conciseness

The body is lean and protocol-specific throughout — e.g., 'Give the three executors distinct approach mandates... conservative/direct, optimized/simplified, and differentiated/ambitious' — with no explanations of concepts Claude already knows; even the version note is framed as archival compatibility, the sanctioned pattern for time-sensitive content.

5 / 5

Actionability

All five verification commands are copy-paste ready with real flags (e.g., 'node <skill-directory>/scripts/verify-evidence.mjs --actor <CRITIC-or-JUDGE> --allow-command <executable> --report <evidence-result.json> <evidence-manifest.json>'), placeholders are genuine runtime parameters, and abstract directives like the stalled-agent policy resolve to required-reading reference files that exist.

5 / 5

Workflow Clarity

The numbered rules, terminal outcomes ('If a lane fails after review three, stop that lane'), feedback loops, and validation at every stage are explicit, but the command block presents 'Then validate the panel decision' before 'Before blind judging, generate the neutral packet', so the tooling sequence is slightly out of chronological pipeline order.

4 / 5

Progressive Disclosure

The body is a clear overview that lists all seven reference files with per-file purpose annotations (e.g., 'references/panel-protocol.md for waves, lane isolation, blind judging, and terminal outcomes'), references are one level deep, and every referenced path resolves to a real file in the bundle.

5 / 5

Total

19

/

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, tightly gated description: it names the concrete panel components, states an explicit and specific invocation condition, and adds a boundary against scope expansion. Its only weaknesses are that the deliverable (a selected winner or no-winner verdict) stays implicit and a few natural synonyms for the trigger concept are absent.

DimensionReasoningScore

Specificity

The description lists several concrete pipeline components — 'three independent executors, one independent paired critic per option, bounded revision loops, and a blind evidence-backed judge' — but leaves the outcome side (winner selection, terminal verdicts) implicit, so coverage has minor gaps rather than being comprehensive.

4 / 5

Completeness

Both what and when are explicitly answered: the what enumerates the panel structure, and the when gives concrete triggers — 'Use only when the user invokes $three-way-judge or explicitly requests this exact panel' — with an added negative scope boundary, matching the top anchor.

5 / 5

Trigger Term Quality

Natural phrasings a user of this niche would say are present ('$three-way-judge', 'three-option workflow', 'this exact panel', 'nine-review deep mode'), but common synonyms such as 'compare three solutions' or 'best of three' are missing, placing it just below comprehensive keyword coverage.

4 / 5

Distinctiveness Conflict Risk

The skill is gated to exact invocation by name, describes 'this exact panel', and explicitly excludes the nine-review deep mode, giving it a clear niche with minimal risk of triggering for the wrong skill.

5 / 5

Total

18

/

20

Passed

Validation

100%

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

Validation — 16 / 16 Passed

Validation for skill structure

No warnings or errors.

Repository

stevenknowswhy/three-critic-judge-skill

Path

SKILL.md

Reviewed

Table of Contents

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.