ablation-planner

Use when main results pass result-to-claim (claim_supported=yes or partial) and ablation studies are needed for paper submission.

Quality

64%

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 ./skills/ablation-planner/SKILL.md

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.

The content is a well-structured, highly actionable workflow with explicit validation and feedback loops for batch ablation runs — a strong skill body. The only meaningful improvement is tightening the illustrative parse template and making execution steps slightly more concrete.

Suggestions

Trim the Step 3 placeholder table rows to one example per section to reduce template length.

Add one concrete example command or config snippet for Step 5's 'Create configs/scripts' to make implementation steps copy-paste ready.

Dimension	Reasoning	Score
Conciseness	The body assumes Claude's ML competence (no 'what is an ablation' primer) and is mostly efficient, but the Step 3 parse template with illustrative placeholder rows ("remove module X", "replace X with simpler Z") is somewhat lengthy and could be trimmed.	4 / 5
Actionability	It provides a concrete Codex MCP invocation with model/config, a fully specified prompt schema (name, what_it_tests, expected_if_component_matters, priority), and concrete file targets (EXPERIMENT_LOG.md), though the actual experiment-execution steps stay high-level ('Run in suggested order').	4 / 5
Workflow Clarity	A clear 5-step sequence (Prepare → Design → Parse → Review Feasibility → Implement) with explicit validation checkpoints (smoke test before full run, feasibility/compute review) and a budget feedback loop ('CC proposes cuts and asks Codex to re-prioritize — don't silently drop ablations') covering batch operations.	5 / 5
Progressive Disclosure	No bundle files exist, so the body is a single self-contained, well-organized file with clear section headers (When to Use, Workflow, Rules) and no nested or broken references; project-file mentions (EXPERIMENT_LOG.md, research_contract.md) are contextual, not skill-bundle refs.	5 / 5
	Total	18 / 20 Passed

Description

43%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 has an excellent, specific trigger condition and a distinct niche, but it omits the 'what' entirely — never stating what the skill actually does — which caps both completeness and specificity. Adding a concrete action clause (e.g., 'Designs and implements ablation studies...') would substantially raise the score.

Suggestions

Add a concrete 'what' clause stating the skill's actions, e.g. 'Designs ablation studies that isolate each novel component, then reviews feasibility and implements them.'

Lead with the capability before the trigger so the description answers 'what does this do' (e.g. 'Designs and runs ML ablation studies for paper submission. Use when...').

Soften internal-only jargon ('claim_supported=yes or partial') by pairing it with natural phrasing a user would actually say.

Dimension	Reasoning	Score
Specificity	The description names the domain ("ablation studies are needed for paper submission") and a precise trigger condition ("claim_supported=yes or partial"), but lists no concrete actions the skill performs — there is no 'designs', 'runs', or 'isolates' verb stating what it does, so actions are minimal/generic.	2 / 5
Completeness	It provides a strong, explicit 'when' ("Use when main results pass result-to-claim...") but no 'what' — the skill's actual capability is never stated, matching the anchor where only 'when' is present without 'what'.	2 / 5
Trigger Term Quality	It includes a couple of natural phrases a user might say ("ablation studies", "paper submission") alongside internal jargon ("result-to-claim", "claim_supported=partial") that users would not naturally utter, leaving common variations and synonyms missing.	3 / 5
Distinctiveness Conflict Risk	It targets a clear, narrow niche tied to a specific research-workflow stage (ablation planning for paper submission after a result-to-claim check), giving distinct triggers with minimal overlap risk.	5 / 5
	Total	12 / 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
allowed_tools_field	'allowed-tools' contains unusual tool name(s)	Warning
frontmatter_unknown_keys	Unknown frontmatter key(s) found; consider removing or moving to metadata	Warning

	Total	14 / 16 Passed

Repository: wanshuiyin/Auto-claude-code-research-in-sleep
Path: skills/ablation-planner/SKILL.md
Commit: f4f20f9

Reviewed: 1 day 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.