Skip to content

dhanoopbhaskar/sdlc-plugins

v0.2.1MIT

Software Development Lifecycle Plugins

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.

[0.2.1] - 2026-11-04

Fixed

  • Evidence grading: behavioral claims now require confirmed evidence, and corroborated behavior is recorded as a hypothesis with a Validation Experiment instead of settled behavior. Process documents are graded by their weakest planning claim, so evidence: confirmed no longer appears alongside unknown or blocked work.
  • OKF output contract: charter.md, ledger.md, and system.md link to sourced concepts instead of carrying sources; parent owns system.md frontmatter, and okf-bundle.md is the single source of truth for section ownership.
  • repo:// evidence: resolution clarified across Git hosts with path percent-encoding, host-specific line-range anchors (GitHub, GitLab, Bitbucket, Azure DevOps), and snapshot-SHA or path-only inferred evidence for dirty working trees.
  • Validation: link checks now verify heading anchors, scanning ignores indented code blocks and HTML comments, version metadata must be semantic versioning, and check_versions / check_actor_versions report non-object JSON and malformed manifests instead of raising AttributeError.
  • Tests: added a minimal OKF bundle fixture and conformance test covering frontmatter, links, repo:// evidence, actors, status values, process-document sources, index/log completeness, and a cross-repository Flow with per-source repo/commit fields.
  • Discovery workflow: skills need read access to targets and write access only to the bundle; flows/ is required only when entry points exist; open questions and risks must be answered and linked; editing a concept clears verified and returns it to status: draft, including when migration-seams adds a parity-hazard section.
  • Orchestration: the worker output contract is depth-aware with system.md fragment returns and parent/worker lease ownership; ledger writes follow a single-writer protocol; added behavior/rules and synthesis phase gates plus a final ledger reconciliation gate.
  • Depth: a small-estate path uses a single inventory shard, while Flows and Integration Boundaries stay deep artifacts with Open Questions at inventory; system-cartography and Step 2 gates are depth-aware so inventory passes are not held to deep requirements. Standalone phases create a minimal charter (partial bundle) and a repos/ registry, or record its absence as an Open Question.
  • Templates: added a # Coupling section owned by migration-seams, an Owner column and atomic groups in system.md sequencing, shard in the risk example, a defined shard Evidence column, and fixture/test-data exclusions.
  • Reconnaissance: commands redact credentials from remote URLs and use grep -r with dependency/build-directory exclusions; added justfile/Taskfile detection, reduced non-git census noise, and recorded only measurable coverage rows while marking later-phase rows not yet measured.
  • Phase descriptions gained trigger-precedence notes so multi-phase requests select the orchestrator, and are clarified as legacy-code-discovery phases.

[0.2.0] - 2026-11-03

Added

  • legacy-code-discovery orchestrator skill for migration-oriented, evidence-graded discovery of legacy systems, with an OKF v0.2 output contract.
  • repo-reconnaissance skill: topology, stack detection, build/run/entry points, deployables, exclusions, and shard planning for large repositories.
  • system-cartography skill: components, typed dependency edges, entry points, flows, integration boundaries, and data ownership.
  • migration-seams skill: risk scoring, coupling and parity analysis, migration options, seams, readiness, and slice sequencing.
  • Discovery reference library: OKF bundle templates, scale and fan-out rules, question bank, detection signals, portable commands, the exploration playbook, and the risk model.
  • Readiness, migration-option, and sequencing sections in the System template, and a parity-hazard variant of the Business Rule template.
  • scripts/validate_skills.py, tests/test_validate_skills.py, and a CI workflow that validate skill frontmatter, naming, links, manifest shape, generated-actor versions, and repository version metadata.

Changed

  • Plugin version bumped to 0.2.0.
  • code-documenter now clears verified when regeneration changes a concept, so human-review metadata cannot outlive the content it verified. Incremental updates preserve verified only when the body and cited sources are unchanged.
  • Discovery skill descriptions were narrowed and given compatibility notes to reduce trigger overlap and clarify co-installation.
  • Read-only, write-target, production-data, and experiment-approval rules were made explicit and consistent across the discovery skills.
  • migration-seams now assesses migration options per deployable and atomic group, with readiness, options, and sequencing recorded in system.md.
  • code-documenter now documents when stable is appropriate; discovery keeps decision-critical concepts at draft until human review.
  • Shard sizing and lease guidance now live in the reference files, with the skills linking to them.

Fixed

  • Reconnaissance: workspace-root detection, census recipes (including untracked and extensionless files), manifest and lockfile coverage, and secret-safe search hygiene.
  • Cartography: dependency-resolution and evidence grading, flow and data-ownership guidance, and the Integration Boundary gate.
  • Bundle and ledger: required index.md/log.md, repo:// evidence conventions, per-source repo/commit for cross-repository citations, worker versus parent ledger ownership, status versus verified, sources only for factual claims, and the local checkout root recorded in the ledger.
  • Templates: status: draft on every concept, <writer>/<version> actors, repos/ and data/ index sections, a # Stack and topology System section, explicit system.md section ownership across phases, and behavior-bearing golden files retained as evidence.
  • Validator: dependency-free frontmatter parsing with quoted/block scalars and duplicate-key, nested-value, multi-line, and tab-indentation rejection; link and image checks including images nested in links, bundle-absolute OKF links, angle-bracket destinations, reference definitions, and tilde/longer-backtick fences; manifest-shape and generated-actor version checks; guarded changelog reads; rejection of empty or whitespace-only compatibility values; and expanded tests.
  • Discovery bundle templates now record repository identity, build/run commands, deployment surface, the full coverage vector, shard notes, and generated-code inventory; standalone phase skills link new concepts and log entries, and standalone resumes verify the charter scope.

[0.1.0] - 2025-09-18

Added

  • Initial project structure