CtrlK
BlogDocsLog inGet started
Tessl Logo
SkillPublic

dspy

DSPy: declarative LM programs, auto-optimize prompts, RAG.

Quality

55%

Does it follow best practices?

Run evals on this skill

Adds up to 20 points to the overall score

View guide
SecuritybySnyk

Low

Low-risk findings worth noting

Fix and improve this skill with Tessl

tessl review fix ./optional-skills/mlops/research/dspy/SKILL.md
SKILL.md
Quality
Evals
Security

Quality

Content

42%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 skill reads as a full DSPy tutorial rather than a lean SKILL.md overview: it is thorough and mostly concrete, but it duplicates its own reference files, pads with explanation Claude doesn't need, and ships several defective code examples. Restructuring around the existing references and fixing the broken snippets would lift it substantially.

Suggestions

Cut the inlined module, optimizer, and example walkthroughs down to one-line pointers to references/modules.md, references/optimizers.md, and references/examples.md, keeping only a short quick-start in the body.

Fix the broken code examples: the unterminated string in the "Optimize with Representative Data" snippet, the no-op isinstance(float(solution), float) assertion, and the deprecated dspy.teleprompt.MIPO / dspy.predict.ReAct import paths.

Remove time-sensitive filler ("GitHub Stars: 22,000+") and either show retriever configuration wherever dspy.Retrieve is used or note it as a prerequisite step.

DimensionReasoningScore

Conciseness

The ~595-line body is noticeably verbose: it explains concepts Claude already knows ("Signatures define the structure of your AI task (inputs → outputs)", "Modules are reusable components that transform inputs to outputs"), includes padding like "**GitHub Stars**: 22,000+ | **Created By**: Stanford NLP", and inlines roughly 200+ lines of module/optimizer/example detail that duplicates the provided references/ files.

2 / 5

Actionability

Most sections give concrete code, but several examples are broken or wrong beyond minor gaps: an unterminated string in the training-data example ("answer="...)"), a logically no-op assertion ("isinstance(float(solution), float)"), deprecated import paths ("from dspy.teleprompt import MIPRO", "from dspy.predict import ReAct"), and "dspy.Retrieve(k=3)" used with no retriever configured.

3 / 5

Workflow Clarity

A coherent progression exists (configure LM → define signature → build module → optimize → evaluate, plus "Start Simple, Iterate"), but validation checkpoints are implicit and scattered across sections rather than integrated into one end-to-end workflow with explicit validate/retry steps.

3 / 5

Progressive Disclosure

The three references (modules.md, optimizers.md, examples.md) are real files, one level deep, and clearly signaled in the "See Also" section, but the body inlines extensive module, optimizer, and example detail that belongs in those separate files — the structure exists yet content placement is inverted.

3 / 5

Total

11

/

20

Passed

Description

53%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 terse and names a distinct, specific framework, so it is unlikely to trigger the wrong skill. But it answers only "what" — there is no "use when" guidance — and its keyword coverage omits the natural synonyms users would actually say.

Suggestions

Add an explicit trigger clause, e.g. "Use when the user mentions DSPy, wants to optimize prompts automatically, or is building RAG pipelines or multi-stage LM programs."

Broaden natural trigger terms with synonyms like "prompt optimization", "Stanford NLP", and "LM pipelines" so the skill surfaces for users who don't name DSPy directly.

DimensionReasoningScore

Specificity

It names the framework and 2-3 actions ("declarative LM programs, auto-optimize prompts, RAG"), but "RAG" is a domain acronym rather than an action and coverage of DSPy's capabilities (optimizers, signatures, evaluation) is thin — matching "names domain and 1-2 concrete actions, but not comprehensive" rather than the several-specific-actions anchor.

3 / 5

Completeness

The "what" is clear (declarative LM programs, auto-optimize prompts, RAG) but there is no "Use when..." clause or equivalent trigger guidance, which per the rubric caps completeness at 3.

3 / 5

Trigger Term Quality

"DSPy", "auto-optimize prompts", and "RAG" are relevant keywords, but common natural variations are missing — users would also say "prompt optimization", "LM pipelines", "Stanford NLP framework", or "few-shot bootstrapping".

3 / 5

Distinctiveness Conflict Risk

Naming "DSPy" explicitly gives it a clear niche with minimal conflict risk; it is not a 5 because the generic trailing terms "auto-optimize prompts, RAG" could overlap with generic prompt-engineering or RAG skills when the DSPy name goes unmentioned.

4 / 5

Total

13

/

20

Passed

Validation

75%

Checks the skill against the spec for correct structure and formatting. All validation checks must pass before discovery and implementation can be scored.

Validation — 12 / 16 Passed

Validation for skill structure

CriteriaDescriptionResult

skill_md_line_count

SKILL.md is long (595 lines); consider splitting into references/ and linking

Warning

metadata_version

'metadata.version' is missing

Warning

metadata_field

'metadata' should map string keys to string values

Warning

frontmatter_unknown_keys

Unknown frontmatter key(s) found; consider removing or moving to metadata

Warning

Total

12

/

16

Passed

Repository

NousResearch/hermes-agent

Path

optional-skills/mlops/research/dspy/SKILL.md

Reviewed

Table of Contents

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.