zero-shot

Use when the user asks to "check zero-shot", "audit few-shot anchoring", "find example anchoring", or invokes /zero-shot. Read-only audit of LLM-facing prose: principle over anchoring examples.

Quality

78%

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 ./epistemic-cooperative/skills/zero-shot/SKILL.md

Quality

Content

67%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.

A well-structured, thoughtful instruction-only skill with concrete scope rules, a sharp boundary test, and a complete output contract. Its weaknesses are density — meta-rationale sections dilute the operating instructions — and the absence of a worked example finding to calibrate the judgment, plus an implicit rather than explicit step sequence.

Suggestions

Add one worked example finding (excerpt + rationale + suggested_rewrite) to the Output section so the judgment has a concrete calibration anchor.

Tighten the Purpose, Distinction, and Confidence sections — they justify the audit's existence rather than instruct it — to cut meta-rationale that competes with the operating instructions.

Render the evaluation flow as an explicit ordered procedure (enumerate in-scope files → evaluate passages outside formal blocks → apply boundary test and exemptions → assign severity → emit JSON) instead of leaving the sequence distributed across prose sections.

Dimension	Reasoning	Score
Conciseness	The core sections (principle statement, boundary test, two exemptions, severity table, JSON schema) earn their tokens as novel domain knowledge Claude does not already have. But Purpose ("a semantic reviewer catches what literal pattern matching cannot"), Distinction, and Confidence carry meta-rationale about the audit's place in a toolchain that could be tightened, matching anchor 3 ("Mostly efficient but includes some unnecessary explanation or could be tightened") rather than anchor 4's minor-trim level.	3 / 5
Actionability	Guidance is mostly executable: explicit in/out-of-scope file lists, a verbatim boundary-test question ("would removing this example increase the LLM's latitude..."), a copy-paste JSON output schema, and a severity-to-surface table. It falls short of anchor 5 because no worked example finding illustrates what a "suggested_rewrite" or rationale actually looks like, which for a judgment-based audit is a real gap.	4 / 5
Workflow Clarity	The sections sequence logically (Inputs → Scope → What to evaluate → Output) and edge-case checkpoints are explicit: ambiguous cases are routed to "severity: low" for human triage, and "When zero findings result, emit the JSON object with empty findings array". Not anchor 5 because the procedure is never laid out as an ordered sequence — the evaluator must assemble the flow from prose sections — but the checkpoints are present, placing it above anchor 3.	4 / 5
Progressive Disclosure	No bundle files exist (references/, scripts/, assets/ are absent), and the body makes no dangling file references; content is single-level and well-sectioned with clear headers. It sits just above the under-50-line exception (the body is ~95 lines) with some arguably peripheral sections (Distinction table, Confidence) inlined, matching anchor 4's "good structure; minor organization gaps" rather than anchor 5.	4 / 5
	Total	15 / 20 Passed

Description

82%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.

A strong description: it explicitly answers both what and when with concrete, niche trigger phrases and poses minimal conflict risk. The only weakness is that a single action ("read-only audit") is named rather than several specific capabilities, leaving coverage of what the audit produces implicit.

Dimension	Reasoning	Score
Specificity	"Read-only audit of LLM-facing prose: principle over anchoring examples" names the domain and one concrete action, but only that single action is enumerated rather than several specific capabilities. Anchor 3 ("Names domain and 1-2 concrete actions, but not comprehensive") is the closest fit; anchor 4 requires several listed actions, which the description does not provide.	3 / 5
Completeness	Both parts are explicit: what ("Read-only audit of LLM-facing prose: principle over anchoring examples") and when ("Use when the user asks to...") with concrete trigger phrases, matching anchor 5 exactly. It is not anchor 4, whose 'when' is only weakly specific — here the when clause carries four concrete triggers.	5 / 5
Trigger Term Quality	Four explicit trigger phrases ("check zero-shot", "audit few-shot anchoring", "find example anchoring", "/zero-shot") give good natural-term coverage with variation. A few natural synonyms are missing (e.g., "remove examples from instructions", "zero-shot instructions"), keeping it below anchor 5's comprehensive synonym coverage; it is clearly above anchor 3's missing-variations level.	4 / 5
Distinctiveness Conflict Risk	"audit few-shot anchoring" and "LLM-facing prose: principle over anchoring examples" define a clear niche with distinct triggers and minimal conflict risk against generic audit or documentation skills. Nothing in the description overlaps with common skill descriptions.	5 / 5
	Total	17 / 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: jongwony/epistemic-protocols
Path: epistemic-cooperative/skills/zero-shot/SKILL.md
Commit: 10abcb3

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.