CtrlK
BlogDocsLog inGet started
Tessl Logo
SkillPublic

adversarial-review

Independently form and test models of what a system should do, challenge the design against them, and work as a review group toward an agreed direction. Use when asking for an adversarial review, an independent architecture review, or an adversarial checkpoint on a plan or cumulative implementation. Do not use for routine final code inspection, which belongs to post-implementation-review.

Quality

88%

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

85%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.

A well-crafted orchestration skill: the workflow is clearly sequenced with explicit verification checkpoints and closure criteria, and the two bundle references are real, well-signaled, and one level deep. The main improvement space is tightening the abstract prose and de-duplicating the anchoring guidance, plus making relay/report formats as concrete as the reviewer prompts.

DimensionReasoningScore

Conciseness

The body is dense with domain-specific guidance and never explains concepts Claude already knows, but the abstract prose style has passages that could be tightened — e.g., anchoring precautions appear both in "Establish independence" ("When anchoring on a proposal is a consequential risk, first supply the outcome…") and again in "Work toward an agreed direction" ("When anchoring is a consequential risk, let it form its own account…"), which could be consolidated. It is not 5 because not every token is maximally lean; it is above 3 because there is no padding or generic explanation.

4 / 5

Actionability

For an instruction-only orchestration skill the guidance is largely executable: a concrete staffing configuration ("Two are fresh Codex subagents… each with `fork_turns: \"none\"` and `reasoning_effort: \"high\"`"), verbatim copy-paste starting questions for each reviewer, an explicit evidence checklist ("Provide the desired outcome, explicit requirements, actual or intended callers, relevant entrypoints, and existing system…"), and explicit closure conditions. Not 5 because some directives remain abstract (e.g., "Reconstruct the work from the desired outcome") and the relay-message and report formats are described but not shown; not 3 because the concrete specifics far exceed pseudocode-level guidance.

4 / 5

Workflow Clarity

The multi-step process is clearly sequenced across four headed phases (independence → form and test → agreed direction → decision and continuation) with explicit validation checkpoints: "Verify deciding claims against actual artifacts", checklist-like closure conditions ("The review closes when every participating reviewer explicitly supports the same concrete next direction, its deciding claims have been checked, and meaningful objections are resolved"), and feedback loops (post-repair checks via post-implementation-review on every repaired file, returning to the group when "repairs materially change reviewed ownership", and escalating product-judgment blockers to the user). This matches the top anchor with validation, error recovery, and closure criteria all explicit.

5 / 5

Progressive Disclosure

Scored against the actual bundle: both referenced files (references/deletion-prizes.md and references/example-workflow.md) exist, are substantive and match their stated purposes, are exactly one level deep, and are clearly signaled with purpose statements ("Read [deletion prizes](references/deletion-prizes.md) for examples where changing a lifetime or operating rule removes coordination"). The core process belongs inline and sibling-skill links (../rethink, ../post-implementation-review, ../consult-claude, ../asymmetric-wins) are appropriate cross-navigation, so structure and navigation match the top anchor.

5 / 5

Total

18

/

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 description: it names concrete actions, gives explicit "Use when" triggers with natural phrasing, and closes with a clear negative boundary against the nearest competing skill. The only improvement space is broader synonym coverage in trigger terms and naming more of the process phases.

DimensionReasoningScore

Specificity

The description lists several concrete actions — "Independently form and test models of what a system should do", "challenge the design against them", "work as a review group toward an agreed direction" — with only minor gaps. It falls short of 5 because later phases of the process (evidence gathering, verification, repair, reporting) are not named, and the action phrasing is somewhat abstract; it exceeds 3 because more than 1-2 specific actions are given.

4 / 5

Completeness

It explicitly answers both questions: the "what" is stated through three concrete actions, and the "when" is explicit with concrete trigger phrases ("Use when asking for an adversarial review, an independent architecture review, or an adversarial checkpoint on a plan or cumulative implementation"). It even adds a negative trigger boundary, so it clearly matches the top anchor rather than the 4 anchor whose "when" could be more specific.

5 / 5

Trigger Term Quality

Natural trigger phrases users would say include "adversarial review", "an independent architecture review", and "adversarial checkpoint on a plan or cumulative implementation". Not 5 because coverage stays within the adversarial/architecture-review family and misses common variations such as "design review", "stress-test", or "red-team"; not 3 because the terms present are ones a user would naturally say rather than jargon.

4 / 5

Distinctiveness Conflict Risk

The niche is distinct (independent adversarial design review) and it explicitly disambiguates from the closest sibling skill: "Do not use for routine final code inspection, which belongs to post-implementation-review". Minimal conflict risk; clearly the top anchor rather than 4, which admits minor overlap.

5 / 5

Total

18

/

20

Passed

Validation

93%

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

CriteriaDescriptionResult

relative_links

Relative link issues: 6 suspicious

Warning

Total

15

/

16

Passed

Repository

EpicenterHQ/epicenter

Path

.agents/skills/adversarial-review/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.