0.5.0

[email protected] | 2026-04-27

Added

• scripts/lib/references.py — reference source, document, load result, and metadata models
• load_local_reference() / load_local_references() — load AGENTS.md from local directories
• load_remote_reference() / load_remote_references() — load AGENTS.md from remote repos via shallow clone
• _run_git() — subprocess helper with 60s timeout and error capture
• scripts/lib/reference_adaptation.py — reference adaptation engine with heuristic classification
• ReferenceSection, AdaptedConvention, ReferenceAdaptationResult — frozen dataclasses for section splitting and convention classification
• split_markdown_sections() — heading-based Markdown section extraction
• adapt_reference() / adapt_references() — classify conventions as applicable, mismatched, uncertain, or ignored
• Category detection with priority ordering: directory_structure, testing, formatter, linter, type_checker, git
• Language/tool extraction and target analysis comparison
• Directory path matching against scan tree
• scripts/lib/reference_questions.py — gap detection and targeted question generation
• ReferenceQuestion model with section, question, reason, category, source, blocking, options
• generate_reference_questions() — produce targeted questions from uncertain and mismatched conventions
• Selective question generation: irrelevant mismatches filtered, ecosystem-aware relevance checks
• Conflict detection across multiple references proposing different conventions
• Question deduplication and deterministic ordering
• scripts/lib/reference_initialization.py — empty-project initialization from references
• is_empty_target() — detect empty or near-empty target repositories
• build_reference_metadata() — build deterministic metadata from loaded documents
• render_reference_metadata_block() — render metadata as Markdown HTML comment with JSON
• ReferenceInitializationResult — structured result with adapted references, questions, metadata, warnings
• initialize_from_references() — end-to-end initialization flow
• successful_reference_documents() — filter load results to successful documents
• AGENTSKILL_VERSION constant

0.4.0

[email protected] | 2026-04-27

Added

• scripts/lib/parsers.py — shared TOML and YAML parser loading with optional dependency fallback
• load_toml / load_yaml — strict parsers that raise ParserUnavailableError when deps are missing
• load_toml_safe / load_yaml_safe — graceful parsers returning {} on any error
• Comprehensive config parsing tests with real-world fixtures for Python, JS, Go, and Rust

Changed

• Replaced custom TOML parsing in config.py with tomllib / tomli via shared parser
• Replaced custom YAML parsing in config.py with PyYAML safe_load via shared parser
• Moved tomli and PyYAML to optional [parsers] dependency group

Removed

• _parse_toml, _parse_toml_value, _split_toml_array — replaced by real TOML parser
• _parse_yaml_simple, _yaml_scalar — replaced by real YAML parser

0.3.0

[email protected] | 2026-04-26

Added

• Logging infrastructure — scripts/lib/logging_utils.py with configurable log level and exception capture
• Output path validation — --out paths are validated; parent directories are created automatically
• Per-analyzer timeout logging — runner logs when an analyzer exceeds its deadline
• write_output exception handling — captures and logs JSON serialization errors

Changed

• _run in git.py now returns stderr alongside stdout, and logs command failures
• run_all logs per-analyzer failures with tracebacks instead of silently swallowing them

0.2.0

[email protected] | 2026-04-26

Added

• GitHub Actions CI workflows — build, test, verify, and main branch checks
• tomli runtime dependency for Python < 3.11 TOML support

Changed

• Bumped minimum Python version to 3.10
• Switched build backend to setuptools.build_meta
• Expanded py-modules to include cli for correct CLI invocation

Fixed

• Import order error flagged by ruff
• CLI invocation error caused by missing py-modules declaration

0.1.0

[email protected] | 2026-04-26

Added

• Initial release of agentskill
• analyze command — run all analyzers and synthesize an AGENTS.md report
• scan analyzer — directory tree mapping, file inventory, suggested read order
• measure analyzer — exact indentation, line length percentiles, blank line distributions
• config analyzer — formatter, linter, and type-checker detection with config excerpts
• git analyzer — commit prefixes, branch naming, merge strategy, and signing detection
• graph analyzer — internal import graph, circular dependencies, most-depended modules
• symbols analyzer — symbol name extraction, naming pattern clustering, affix detection
• tests analyzer — test-to-source mapping, framework detection, fixture extraction
• Parallel analyzer execution via ThreadPoolExecutor
• Pretty-printed and machine-readable JSON output modes (--pretty, --json)
• --out flag to write report to a file
• --language flag to override auto-detected language
• Language-agnostic analysis engine supporting Python, JavaScript/TypeScript, Rust, Go, and others
• Multiple output example formats: SINGLE_LANGUAGE.md, MULTI_LANGUAGE.md, MONOREPO.md
• SYSTEM.md — behavioral spec for the synthesis step
• SKILL.md — OpenClaw AgentSkill manifest
• AGENTS.md — self-documented analysis rules
• Full test suite with pytest covering all modules
• pyproject.toml with project.scripts entry points
• Development dependencies: ruff, pytest