autoresearch

Run bounded automated experiment iterations by recording baselines, applying hypothesis patches, comparing metrics, protecting regression guards, and deciding keep, discard, rollback, or block. Use when automated research is requested or a repo/skill needs evidence-backed research, metric tracking, or safe optimisation loops.

Quality

—

Does it follow best practices?

Impact

—

No eval scenarios have been run

Securityby

Advisory

Suggest reviewing before use

Quality

Content

85%

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

The body is well-structured, actionable, and uses progressive disclosure effectively with real reference files. The main improvement opportunity is reducing thematic redundancy across the Constraints, Decision Language, Gotchas, and Anti-Patterns sections.

Suggestions

Consolidate the repeated destructive-command refusal and guard-regression rules into one authoritative section to remove redundancy between Constraints, Decision Language, Gotchas, and Anti-Patterns.

Consider folding the short Avoid and Anti-Patterns sections together, since both enumerate what not to do, to further tighten the document.

Dimension	Reasoning	Score
Conciseness	Sections are terse and assume Claude's competence (no concept padding), but themes recur across Constraints, Decision Language, Gotchas, and Anti-Patterns (e.g. destructive-command refusal and guard-regression rules appear multiple times) and could be tightened. Not level 3 because of this redundancy; not level 1 because there is no explanatory fluff.	2 / 3
Actionability	Provides executable, copy-paste-ready commands ("uv run train.py --steps 200 --json", "./bin/ask skills audit … --json --robot") and a concrete YAML ledger entry. Above level 2, which would offer only pseudocode or incomplete guidance.	3 / 3
Workflow Clarity	The 9-step Workflow is clearly sequenced with explicit validation gates ("Baseline first. Never keep an experiment before baseline evidence exists", Verify/Guard before keep/discard) and a Repair feedback loop (rerun the failing gate first). Appropriate for destructive/batch experiment operations, so above level 2.	3 / 3
Progressive Disclosure	An explicit "Progressive Disclosure" section points one level deep to references/autoresearch-project.md, contract.yaml, evals.yaml, task-profile.json, and discovery-interview.md — all of which exist as real files. Clear overview with well-signaled navigation, matching level 3.	3 / 3
	Total	11 / 12 Passed

Description

85%

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, specific third-person description that names concrete actions and gives explicit trigger guidance. The only weakness is trigger-term naturalness, where some phrasing is technical rather than what a user would spontaneously say.

Suggestions

Soften trigger phrasing toward natural user language, e.g. add "run experiments", "automated research", or "iterate on a repo with metrics" alongside the technical terms.

Consider including the bare skill name "autoresearch" in the description trigger clause so it matches users who say the word directly.

Dimension	Reasoning	Score
Specificity	Lists multiple specific concrete actions: "recording baselines, applying hypothesis patches, comparing metrics, protecting regression guards, and deciding keep, discard, rollback, or block." This matches the level-3 anchor rather than level 2, which would only name a domain and some actions.	3 / 3
Completeness	Explicitly answers both what ("Run bounded automated experiment iterations by recording baselines…") and when ("Use when automated research is requested or a repo/skill needs evidence-backed research…"). Not level 2 because the "when" is explicit rather than merely implied.	3 / 3
Trigger Term Quality	Trigger phrases like "automated research is requested" are natural, but "evidence-backed research, metric tracking, or safe optimisation loops" lean technical and miss common user variations. It is above level 1 (which has no natural keywords) but lacks the broad natural-term coverage of level 3.	2 / 3
Distinctiveness Conflict Risk	Occupies a clear niche (bounded, ledger-backed experiment loops with regression guards) with distinct triggers unlikely to fire for unrelated skills. Above level 2, which would still risk overlap with similar skills.	3 / 3
	Total	11 / 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
metadata_field	'metadata' should map string keys to string values	Warning

	Total	15 / 16 Passed

Repository: jscraik/Agent-Skills
Commit: 5a6027f

Reviewed: 5 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.