CheckGroup identifies a category of checks that can be enabled or disabled.

Level represents the severity of a validation result.

AllowedTools handles the type ambiguity in the allowed-tools field. The spec defines it as a space-delimited string, but many skills use a YAML list i

CachedResult holds a scoring result with metadata for cache storage.

ClientOptions holds configuration for creating an LLM client.

ContaminationReport holds cross-language contamination metrics.

ContentReport holds content quality metrics computed by the content analyzer.

DimensionScore holds a single scoring dimension's display name and value.

--- JSON output --- EvalJSONOutput is the top-level JSON envelope.

EvalJSONRef is one reference file entry in JSON output.

EvalJSONSkill is one skill entry in JSON output.

Frontmatter represents the parsed YAML frontmatter of a SKILL.md file.

LLMClient is the interface for making LLM API calls.

MultiReport holds aggregated results from validating multiple skills.

Options controls what gets scored.

Options configures which checks Validate runs.

Options controls which checks RunAllChecks performs.

ProgressFunc receives progress events during evaluation. event identifies the kind of event (e.g. "scoring", "cached", "warning", "error"). detail pro

RefResult holds scoring output for a single reference file.

RefScores holds the LLM judge scores for a reference file.

ReferenceFileReport holds per-file content and contamination analysis for a single reference file.

Report holds all validation results and token counts.

Result represents a single validation finding.

Result holds the complete scoring output for one skill.

ResultContext is a builder that sets Category and a default File once, so individual results inherit location context automatically.

Scored is the interface implemented by both SkillScores and RefScores. It allows formatting code to iterate dimensions generically.

Skill represents a parsed skill package.

SkillMode indicates what kind of skill directory was detected.

SkillScores holds the LLM judge scores for a SKILL.md file.

TokenCount holds the token count for a single file.

--- Anthropic client ---

--- capturingMockClient records arguments for each call ---

--- Claude CLI client --- claudeCLIClient invokes the "claude" CLI for completions. This is useful when the CLI is already authenticated (e.g. via a c

exitCodeError is a sentinel error that carries a non-zero exit code. It is returned by output helpers and handled by Execute().

exitOpts controls how validation results map to exit codes.

--- Mock client for integration-style tests ---

--- Mock LLM client ---

--- OpenAI-compatible client ---

queueItem represents a text body to scan during the BFS reachability walk.

rewriteTransport rewrites requests to a different target URL while preserving the original Host header for testing.

--- sequentialMockClient ---