Tessl • The management layer for AI agent skills

Blog Docs Log inGet started

SkillPublic

do-and-judge

Execute a task with sub-agent implementation and LLM-as-a-judge verification with automatic retry loop

Quality

70%

Does it follow best practices?

Run evals on this skill

Adds up to 20 points to the overall score

View guide

Securityby

Low

Low-risk findings worth noting

Fix and improve this skill with Tessl

tessl review fix ./plugins/sadd/skills/do-and-judge/SKILL.md

Quality

Content

70%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 thorough, well-sequenced orchestration workflow with strong validation and feedback loops, but it is notably verbose and monolithic. Breaking the long examples and reference tables into separate files would substantially improve token efficiency and navigation.

Suggestions

Move the five full worked examples into a separate references/examples.md file and link to it, keeping only a brief Quickstart example inline.

Extract the Model Selection Policy tables and Cross-Provider Equivalence map into a references/model-selection.md file referenced from SKILL.md, reducing the main body's token load.

Tighten or remove the exclamation-heavy threats (e.g., "you will be killed immediatly!!!!") and fix typos such as "sadd:meta-judge" → "sdd:meta-judge", "imidiatly", "sumaraize", and "coralation".

Dimension	Reasoning	Score
Conciseness	Mostly efficient in laying out the orchestration, but several sections are padded — repeated escalation/role-pairing guidance, exclamation-laden threats ("you will be killed immediatly!!!!"), and five very long inlined examples — that could be trimmed without losing clarity.	3 / 5
Actionability	Provides concrete, copy-paste-ready prompt templates, dispatch examples, and structured decision-logic blocks; minor gaps come from typos like "sadd:meta-judge"/"sadd:judge" (vs. the referenced "sdd:" agents), "imidiatly", and "coralation".	4 / 5
Workflow Clarity	The six-phase process is explicitly sequenced with validation checkpoints (judge verification between attempts), clear feedback loops (parse verdict → fix → retry), retry-budget reset rules, and explicit max-retry escalation.	5 / 5
Progressive Disclosure	There is section-level structure (phases, headers), but with no bundle files present everything is inlined into a ~1240-line monolith — the lengthy model-selection tables, prompt templates, and five full examples clearly belong in separate reference files.	3 / 5
	Total	15 / 20 Passed

Description

58%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 states a concrete, mostly-distinct capability but omits any explicit "when to use" trigger guidance, which caps its completeness. Trigger terms are relevant but technical rather than natural user phrasing.

Suggestions

Add an explicit 'Use when...' clause stating the natural situations that call for this skill (e.g., 'Use when the user wants a task done with automated quality verification, or when a single-shot attempt is too risky').

Include natural trigger phrases and synonyms a user would actually say (e.g., 'run a task and verify it', 'do this with a reviewer', 'implement and check') rather than only technical terms like 'LLM-as-a-judge'.

Tighten the generic 'Execute a task' opener to a more specific verb set so the skill is less likely to overlap with general coding assistants.

Dimension	Reasoning	Score
Specificity	Lists several concrete actions — "sub-agent implementation", "LLM-as-a-judge verification", "automatic retry loop" — naming the domain plus distinct operations, but stops short of the comprehensive multi-action coverage of the score-5 anchor.	4 / 5
Completeness	Gives a clear "what" (execute with sub-agent + judge verification + retry) but the "when" is entirely absent with no "Use when..." clause, which per the judging guidelines caps completeness at 3.	3 / 5
Trigger Term Quality	Contains relevant keywords ("Execute a task", "LLM-as-a-judge", "retry loop") but leans technical and omits natural phrases a user would actually say; no synonyms or extensions are covered.	3 / 5
Distinctiveness Conflict Risk	The sub-agent + judge + retry-loop combination is a fairly distinct niche with minimal overlap risk; the generic "Execute a task" prefix is the only source of minor overlap with general coding skills.	4 / 5
	Total	14 / 20 Passed

Validation

87%

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 — 14 / 16 Passed

Validation for skill structure

Criteria	Description	Result
skill_md_line_count	SKILL.md is long (1242 lines); consider splitting into references/ and linking	Warning
frontmatter_unknown_keys	Unknown frontmatter key(s) found; consider removing or moving to metadata	Warning

	Total	14 / 16 Passed

Repository: NeoLabHQ/context-engineering-kit
Path: plugins/sadd/skills/do-and-judge/SKILL.md
Revision: 23e2428

Reviewed: 26 days ago

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.

do-and-judge

Table of Contents

Is this your skill?