training-stack-change

Modifies training-stack workflows, selector or embedding pipelines, evaluation artifacts, or runtime-facing outputs under src/training. Use when changing model selection training, embedding pipelines, evaluation scripts, experiments, or other training outputs that feed runtime behavior.

Quality

75%

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 ./tools/agent/skills/training-stack-change/SKILL.md

Quality

Content

75%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 clean, well-organized ~30-line skill with concrete verification commands and two validation gates in its workflow. It sits at the 4-anchor throughout, held from 5 by a few abstract workflow steps and the absence of an explicit validation-failure feedback loop.

Suggestions

Make abstract workflow steps more concrete: replace 'Modify the training workflow, artifact expectation, or runtime-facing output' and 'Record any runtime-facing artifact contract mismatches as indexed debt entries' with the specific file/entry operations to perform.

Add an explicit feedback loop after the validation gates, e.g. 'If agent-ci-gate fails: review the reported constraint, fix in the changed surface, and re-run agent-ci-gate before proceeding.'

Tighten the Gotchas wording (e.g. 'do not force selector, embedding, eval, or experiment work through classifier language that hides the real contract') to a shorter, more directive form.

Dimension	Reasoning	Score
Conciseness	The body is lean with short, non-padded sections and no concept-explanation fluff, assuming Claude's competence; a few phrasings in Gotchas ('do not force selector, embedding, eval, or experiment work through classifier language that hides the real contract') could be tightened slightly.	4 / 5
Actionability	Provides concrete executable commands ('make agent-report ENV=cpu CHANGED_FILES="..."', 'make agent-ci-gate CHANGED_FILES="..."') but several workflow steps (1, 2, 5) remain high-level and abstract ('Modify the training workflow...', 'Record any runtime-facing artifact contract mismatches as indexed debt entries').	4 / 5
Workflow Clarity	A clear 5-step sequence with two explicit validation checkpoints (agent-report verifies surface alignment, agent-ci-gate validates constraints), but no explicit error-recovery feedback loop ('if validation fails, fix and re-run') is spelled out for these batch/runtime-facing operations.	4 / 5
Progressive Disclosure	Well-organized sections (Trigger, Workflow, Gotchas, Must Read, Standard Commands, Acceptance) with clearly signaled one-level-deep references to real docs (change-surfaces.md, repo-map.md); no bundle files exist so the skill is appropriately self-contained, with only minor organization gaps.	4 / 5
	Total	16 / 20 Passed

Description

75%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 is well-structured with an explicit 'Use when' trigger clause and a clearly scoped domain under src/training, hitting the 4-anchor across all dimensions. It falls short of 5 mainly because 'Modifies' is a single broad verb applied to many objects rather than a list of distinct concrete actions.

Dimension	Reasoning	Score
Specificity	Names the training-stack domain and several concrete action surfaces ('workflows, selector or embedding pipelines, evaluation artifacts, or runtime-facing outputs'), but the actions all collapse to the single verb 'Modifies' rather than multiple distinct concrete actions, keeping it below the top anchor.	4 / 5
Completeness	Explicitly answers both 'what' ('Modifies training-stack workflows... under src/training') and 'when' ('Use when changing...'), but the 'what' is a single broad verb rather than the multiple concrete actions the top anchor expects.	4 / 5
Trigger Term Quality	The 'Use when' clause lists natural developer terms ('changing model selection training, embedding pipelines, evaluation scripts, experiments'), but lacks file extensions and common synonyms and leans on jargon like 'runtime-facing outputs'.	4 / 5
Distinctiveness Conflict Risk	Scoped to src/training with distinct triggers (training-stack, selector, embedding, evaluation, experiments) giving a clear niche, though 'runtime-facing outputs' introduces some breadth that slightly raises overlap risk.	4 / 5
	Total	16 / 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
frontmatter_unknown_keys	Unknown frontmatter key(s) found; consider removing or moving to metadata	Warning
relative_links	Relative link issues: 2 suspicious	Warning

	Total	14 / 16 Passed

Repository: vllm-project/semantic-router
Path: tools/agent/skills/training-stack-change/SKILL.md
Commit: e73c772

Reviewed: 9 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.