experiment

Use when a quest is ready for a concrete implementation pass or a main experiment run tied to a selected idea and an accepted baseline.

Quality

60%

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 ./src/skills/experiment/SKILL.md

Quality

Content

62%

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

The content is a well-structured control surface with a clear sequenced workflow, explicit validation gates, and properly one-level-deep references that hold real content. Its main weaknesses are length and duplication across overlapping gate sections, plus a mix of concrete hooks with abstract directive prose.

Suggestions

Collapse the overlapping Validation, Acceptance gate, Non-negotiable rules, and Run-quality rules sections into a single ordered gate, and delete the verbatim duplicate lines (41-44) to recover token budget.

Push the evidence-ladder detail and failure-type taxonomy (lines 167-177, 240-258) into a reference file, leaving only the decision rule inline, to improve progressive disclosure.

Tighten abstract directives (e.g. "Run only the checks needed to maximize valid evidence per unit time") into concrete, actionable guidance or move them to the execution playbook.

Dimension	Reasoning	Score
Conciseness	The body is mostly efficient and assumes Claude's competence with no basic-concept padding, but at ~269 lines it carries real duplication — e.g. lines 41-44 restate the PLAN.md and extra-metrics rules verbatim, and the Validation, Acceptance gate, Non-negotiable rules, and Run-quality rules sections overlap heavily.	2 / 3
Actionability	Concrete callable hooks are present (artifact.record_main_experiment(...), bash_exec(...), artifact.git(...), memory.list_recent(...)) alongside concrete file conventions, but much of the guidance is abstract directive prose ("Run only the checks needed to maximize valid evidence per unit time") rather than copy-paste-ready steps.	2 / 3
Workflow Clarity	The numbered Control workflow gives a clear sequence (lock contract, implement, smoke/pilot, execute and monitor, validate and record) with explicit validation checkpoints (Validation, Acceptance gate) and feedback loops for error recovery (diagnosis mode, last-known-good, isolate-and-retry).	3 / 3
Progressive Disclosure	References are one level deep and well-signaled with a "Use them when" trigger (plan/checklist templates, execution-playbook, operational-guidance), and each referenced file is self-contained, but the body itself stays monolithic with evidence-ladder, failure taxonomies, and run-quality rules inlined rather than pushed out.	2 / 3
	Total	9 / 12 Passed

Description

57%

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 has an explicit, well-qualified trigger with a clear niche, but it states the trigger condition more concretely than the skill's actual capabilities and leans on internal framework jargon. Tightening it to name the concrete actions (turn a selected route into one measured result) alongside the trigger would lift specificity and completeness.

Suggestions

Lead with a concrete capability statement (e.g., "Turn one selected route into one trustworthy measured result...") before the "Use when" clause so the "what" is explicit, not just implied by the trigger.

Pair the internal terms ("selected idea", "accepted baseline") with more natural phrasings a user might actually say (e.g., "main experiment run", "measured comparison against baseline") to improve trigger-term quality.

Add one or two concrete actions the skill performs (e.g., "records the run durably and routes from the measured evidence") to raise specificity from domain-naming to multi-action.

Dimension	Reasoning	Score
Specificity	The description names the domain and the kind of pass ("a concrete implementation pass or a main experiment run") but does not enumerate multiple specific concrete actions the skill performs, sitting between the domain-plus-actions anchor and the multi-action anchor.	2 / 3
Completeness	The "when" is explicit and well-qualified ("Use when a quest is ready... tied to a selected idea and an accepted baseline"), but the "what" is only implied through the trigger condition rather than stated as a concrete capability, so what-and-when is only partially covered.	2 / 3
Trigger Term Quality	It opens with a natural "Use when" trigger ("a quest is ready for a concrete implementation pass or a main experiment run"), but terms like "selected idea" and "accepted baseline" are internal quest-framework taxonomy rather than vocabulary a user would naturally say.	2 / 3
Distinctiveness Conflict Risk	The "selected idea and an accepted baseline" qualifiers carve a clear niche that distinguishes it from framing, route-selection, and writing skills, making wrong-skill triggering unlikely.	3 / 3
	Total	9 / 12 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
frontmatter_unknown_keys	Unknown frontmatter key(s) found; consider removing or moving to metadata	Warning

	Total	15 / 16 Passed

Repository: ResearAI/DeepScientist
Path: src/skills/experiment/SKILL.md
Commit: b366244

Reviewed: 2 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.