Skip to content

CHANGELOG.md

Latest commit

 

History

History
35 lines (25 loc) · 5.21 KB

CHANGELOG.md

File metadata and controls

35 lines (25 loc) · 5.21 KB
Raw

Changelog

All notable changes to The Agentic SDLC Handbook are recorded here. Format follows the spirit of Keep a Changelog; the handbook is prose, so entries describe edits to chapters and primitives rather than software diffs.

0.10.1 - May 2026 edition

Reference architecture (Ch04)

  • New Ch04 ("The Reference Architecture") presents a 5-layer stacked supply-chain landscape: Platform -> Context & Capabilities -> Governance and Distribution -> Agent Harness -> SDLC phases. Bottom-to-top hero diagram with per-edge labels (hosts / authored as / resolves and ships / executes as).
  • Recursion diagram: the {Skill, Persona, Context} triplet repeats at every depth; eval-and-plan conventions make the recursion governable.
  • APM-as-supply-chain table maps apm.yml, apm.lock.yaml, apm_modules/, apm install, apm publish to their npm analogues.
  • Frameworks-as-libraries sidebar names Spec-Kit, Squad, Superpowers, and BMAD as primitive-supplying libraries at the Context & Capabilities layer (with URLs).
  • Closing thesis: the SDLC is one instance of the pattern; legal review, M&A diligence, financial close, marketing campaigns, and book authoring all generalise the same way. This handbook itself was authored using the editorial-pipeline Skill (forward-pointer to case-study-handbook-writing.qmd in Part IV).

Part III restructure (Ch08-Ch21)

  • Reorganized Part II Ch04 ("The Reference Architecture") as a leader-grade chapter: surface 5-layer landscape, decision matrix, build/buy/compose framing — without the deep recursion mechanics, panel walkthrough, or three-concern triplet that previously lived alongside the executive material.
  • Promoted the practitioner-grade material into a new Ch08 ("The Practitioner's Reference Architecture") at the head of Part III: deep mechanics of the five-layer model, end-to-end Panel walkthrough (Section 4), recursion bound, and the start-anywhere pattern generalisation — all with -mechanics-suffixed labels (e.g. @sec-five-layers-mechanics, @fig-recursion-mechanics) so they do not collide with the leader-grade Ch04 anchors.
  • Renumbered the remaining Part III chapters by +1: Ch08 Mindset -> Ch09, Ch09 Runtime -> Ch10, Ch10 Instrumented -> Ch11, Ch11 PROSE -> Ch12, Ch12 Load Lifecycle -> Ch13, Ch13 Attention -> Ch14, Ch14 Det/Prob Boundary -> Ch15, Ch15 Multi-Agent -> Ch16, Ch16 Execution Meta-Process -> Ch17, Ch17 Rosetta -> Ch18, Ch18 Anti-Patterns -> Ch19, Ch19 Primitives-as-Code -> Ch20, Ch20 What Comes Next -> Ch21.
  • Updated _quarto.yml Part III chapter list and Closing block to reflect the new ordering.
  • Swept all Chapter N and ChNN prose mentions across every chapter to point at the new numbering. Refs to deep mechanics (Panel deep treatment, recursion bound, three-concern triplet, npm-mapping) re-pointed from old Ch04 to new Ch08; refs to surface-level material (5-layer landscape mental model, surface starter shape) stayed at Ch04. Rewrote the Part III opener forward-pointer in Ch07 to reflect that Ch08 now leads with the practitioner-grade reference architecture (not the mindset chapter, which moved to Ch09).

Roles and team structures (Ch06)

  • Three role concerns named: Domain Specialist (procedure), Agentic Workflow Engineer (composition), Agent Operations Specialist (operations). The BPM workflow-engineer collision is named once and dispensed with. SME and Domain Owner are recognised passing synonyms for Domain Specialist.
  • 5-layer ownership swimlane shows who builds and operates what at each layer.
  • New subsection "How developers work in the agentic SDLC" treats the developer day-to-day: authoring Skills, Personas, and Context bundles in the same repo as the code; CODEOWNERS extends to apm.yml; inner-loop / outer-loop split.

Cross-cutting

  • Compounding and cost-curve claims throughout the book carry hypothesis or early signal register.
  • All chapter diagrams use the centralised _quarto.yml mermaid CSS variables (lines 76-100). Per-diagram colour skinning is not used.
  • 8 sibling chapters (Ch03, Ch05, Ch13, Ch15, Ch17, Ch19, Ch20) refreshed for cross-link fidelity and 5-layer vocabulary alignment.
  • Ch20 closing thread: "the SDLC is just one example of what these patterns and harnesses can agentify."
  • Split Part III opening: a short preface (vocabulary + 5-layer map + composition-pattern names, ~1,600 words) opens Part III as the new Ch08 ("A Map for Part III"), giving the leader a five-minute exit ramp and the practitioner the vocabulary the substrate chapters operate in. The capstone "The Reference Architecture, Earned" (Ch21) closes Part III with the Panel worked example, the APM-to-npm mapping table, the Monday decision frame, and the "the proof is this book" close — landing as earned recognition after the substrate chapters rather than as an asserted preview before them. Synthesis-grade material moved from the opener to the capstone; vocabulary-introduction material stayed at the opener. Ch07's Part-III forward pointer, Ch09's opening callback, and the four substantive Ch08 cross-references in Ch14, Ch16, and Ch20 were re-pointed accordingly. Closing chapter "What Comes Next" renumbered from Ch21 to Ch22. Two legacy footnote names in Ch09 (^ch8-n1, ^ch8-cognitive) renamed to ^ch9-n1, ^ch9-cognitive.