Changelog

All notable changes to this project will be documented in this file.

The format is based on Keep a Changelog, and this project adheres to Semantic Versioning.

[Unreleased]

Changed

  • adr-author (v1.0.0 → v1.1.0) - Adds scoping guidance, corrects the supersession rule, and corrects one point of Nygard attribution.
    • New “Scoping an ADR” section. The skill said when an ADR is warranted but nothing about how many. The test given is supersession granularity: an ADR’s value is that it can be pointed at and reversed on its own, so bundling couples the futures of the decisions bundled. Includes the observation that an over-broad draft is usually not N ADRs but two decisions that never warranted a record, plus where those belong instead.
    • Supersession corrected. The skill said never to rewrite an accepted ADR. Nygard says the opposite about the status and the forward reference: “if a decision is reversed, we will keep the old one around, but mark it as superseded”, “with a reference to its replacement”. Editing the old record to point at its replacement is the mechanism. What must not be rewritten is the Context, Decision, or Consequences.
    • Partial correction. deprecated and superseded both retire a whole record, and the format offers nothing for “still in force except for one claim” — a case that arises as soon as ADRs start correcting each other in detail. Documents the convention this repository already uses in docs/adr/0002-federated-protocol-repositories.md: name the amended part in the Status line, and explain it in a banner under the metadata block.
    • Attribution. Alternatives considered was labelled part of the Nygard format. Nygard’s elements are Title, Context, Decision, Status, Consequences. The section is kept, now named as the later addition it is.

Removed

  • bioc-protocol-runner - Moved to waldronlab/agent-protocol-standard and renamed protocol-runner. Its behaviour — trust ranking, protocol status handling, the PROTOCOLS.yaml fields it reads, and the Method Provenance block — is defined by PROTOCOL_STANDARD.md rather than by Bioconductor, so it now versions with that standard. Anyone who installed it should repoint at agent-protocol-standard/skills/protocol-runner. Amends ADR 0002 §1.

Added

  • bioc-pkg-dev (v1.0.0) - Guides an R package or a set of analysis scripts through Bioconductor submission: the pre-submission gate, the Contributions tracker, and post-acceptance maintenance.
    • Reports gate items in two tiers, separating what upstream states as requirements from what it states as recommendations, so a recommendation is not reported as a blocker.
    • Ships 21 task-oriented knowledge/ summaries of the official guide, each linking back to its canonical chapter, loaded per topic rather than on every invocation.
    • Drives BiocCheck and biocthis rather than reimplementing validation or templates.
  • SKILLS.md and USAGE.md entries for the new skill.

[3.0.0] - 2026-07-07

Added

  • CI/CD Skills Validation
    • Added GitHub Actions workflow .github/workflows/validate-skills.yml to run structural validation automatically on PRs.
    • Added deterministic check script scripts/validate_skills.py to enforce required frontmatter fields, name matching, uniqueness, link integrity, and version bumps.
    • Added .github/copilot-instructions.md to guide GitHub Copilot Code Review for inline subjective feedback on PRs.
  • Bioc-HowTo Integration
    • Merged waldronlab/ai-agent-skills into the official bioconductor/ai-agent-skills repository.
    • Added bioc-howto skill and 11 R-oriented how-to guides (e.g. coverage computation, sequence composition, GRanges manipulation, BAM reading).
  • GitHub Pages Navigation
    • Improved documentation structure and FAQ accessibility on GitHub Pages.

Changed

  • validate-skill (v1.1.0 → v1.2.0) - Updated with CI/CD integration note clarifying the local vs CI/CD division of labor.
  • antigravity adapter (formerly gemini) - Updated setup instructions and renamed adapter to instructions/antigravity.md.

[2.1.0] - 2026-04-03

Added

  • security-audit-r-package (v1.0.0) - Comprehensive security auditing for R/Bioconductor packages
    • Analyzes R code, native code (C/C++/Fortran), dependencies, and package structure
    • References external gist as single source of truth for security checks and issue labels
    • Flexible output formatting (structured reports for full audits, concise for quick checks)
    • Supports targeted audits (specific files/directories) or full package scope
  • AGENTS.md § Avoiding Content Duplication - Guidelines for preventing SSOT violations
  • SKILL_STANDARD.md § SSOT Compliance checklist

Changed

  • create-skill (v1.2.1 → v1.3.0) - Added comprehensive “Avoiding SSOT Violations” guidance
    • SSOT = Single Source of Truth (avoid duplicating content across skills, documentation, validation)
    • Lists what not to duplicate (YAML specs, validation criteria, external checklists)
    • Provides reference patterns with examples
    • Documents signs of SSOT violation and fix patterns
  • validate-skill (v1.0.1 → v1.1.0) - Added SSOT violation detection
    • New validation step 8: “Check for SSOT Violations”
    • Flags common duplication patterns (field lists, templates, copied checklists)
    • Validates proper reference patterns
  • document-skill (v1.0.1) - Reduced duplication in YAML field descriptions

Improved

  • Meta skills now reference AGENTS.md as single source of truth (~95 lines of duplication removed)
  • All YAML field specifications reference AGENTS.md § Skill File Format
  • All validation criteria reference AGENTS.md § Minimal Required Fields / What NOT to Include
  • Validation report templates simplified (reduced from 76 lines to ~30 lines)
  • SKILL_STANDARD.md updated to v2.1.0 with SSOT compliance checklist

[2.0.0] - 2026-04-03

Changed

  • BREAKING: Migrated to agent-agnostic skill format
  • Removed platforms: and triggers: fields from all skills
  • Shifted to natural language invocation via SKILLS.md discovery
  • Updated all skills to use platform-agnostic language
  • Platform shortcuts now documented only in instructions/{agent}.md

Added

  • AGENTS.md - Canonical agent behavior standard
  • SKILL_STANDARD.md - Quick reference with checklist and examples
  • document-skill - Automates SKILLS.md updates after skill creation
  • validate-skill - Validates skills against repository standards
  • check-bioconductor-skills - Verifies installation and lists available skills
  • Complete r-packages skill suite (analyze-r-package, create-package-instructions, update-package-instructions)
  • instructions/ directory with platform adapters (claude.md, copilot.md, antigravity.md)
  • Migration guide in SKILL_STANDARD.md for v1.x users

Improved

  • SKILLS.md now serves as single source of truth for skill catalog
  • Comprehensive documentation with clear separation of concerns
  • Example from parkinsonsMetagenomicData (data-package-parquet)

[0.1.0] - 2026-04-03

Added

  • Initial release
  • Repository structure created
  • Basic documentation framework
  • First experimental skills

Version Guidelines

Major Version (X.0.0)

  • Breaking changes to skill interface
  • Incompatible instruction format changes
  • Major reorganization

Minor Version (0.X.0)

  • New skills added
  • New examples added
  • New features in existing skills (backwards compatible)
  • New domain directories

Patch Version (0.0.X)

  • Bug fixes
  • Documentation improvements
  • Template refinements
  • Example updates
This site uses Just the Docs, a documentation theme for Jekyll.