Changelog

Tease: Track what changed across the skills, shared packaging, and public site.
Lede: This changelog records user-visible collection changes with dates and verification boundaries.
Why it matters: A growing collection needs one place to distinguish skill behavior, installation metadata, and documentation releases.
Go deeper: Follow the dated entries and linked verification pages for the evidence behind each release claim.

2026-07-17

Relocated the repository-wide skill and instruction system audit and its remaining roadmap to Harness Atlas, while keeping the HTMA behavior contract in this collection.
Strengthened HTMA Measure so missing value-of-information inputs remain unknown or explicitly qualitative instead of receiving numerical defaults.
Added the VOI score state, nullable expected value, and missing-input contract to the feature and architecture documentation.
Recorded the remaining audit sequence on the project roadmap.

Added a source-backed guide to documented estimation pains, including budget-before-quote decisions, point-estimate risk, hidden work, target-price ambiguity, small samples, research stopping rules, and asymmetric error costs.
Added copyable prompts that use explicit bracketed fields rather than fabricated measurements or realistic-looking placeholder results.
Added a human-facing README beside the canonical HTMA Measure skill and aligned the root repository guide and Starlight navigation.

Ported the original generic HTMA scoring utilities and scrubbed aggregate H1-H12 history beside the skill.
Added a transparent ten-case H12-derived regression packet, deterministic scorer, isolated Claude CLI runner, and recorded native-agent baseline.
Used Zach Prompting to reconcile estimate-versus-block transitions, then compared the unchanged skill and two isolated candidates on the same regression suite.
Added an explicitly synthetic six-case H13 acceptance suite for missing-threshold, current-lookup, ambiguous-target, and unavailable-private-actual behavior.
Updated HTMA Measure so a missing threshold blocks only the action comparison when a numeric range remains responsible, while missing identifiers, current sources, and private actuals still block numeric output.
Preserved the private boundary for source ledgers, raw historical outputs, mappings, human score sheets, and active sealed holdouts.
Reclassified the sanitized H12 packet as a visible regression set; its results are not fresh holdout evidence.

Reframed the repository and documentation site as the collection of skills Sam uses and shares.
Added collection-level About and Installation pages, and moved collection navigation ahead of skill-specific sections.
Changed the bundled Claude Code namespace from htma-measure to sam while retaining the stable marketplace install key.
Verified public commit 8405b24 through a fresh HTTPS marketplace install, strict manifest validation, component inventory, and recursive skill-content comparison.

Added the standalone zach-prompting skill for reviewing, rewriting, debugging, and migrating prompts and agent instruction artifacts.
Added conditional GPT-5.6 and Claude Fable 5 guidance sourced from the official OpenAI and Anthropic prompting guides.
Added four fresh-context forward uses covering rewrite, review-only, model migration, and a real repository instruction audit; recorded their scope and limitations on the verification page.
Added the public Zach Prompting guide, examples, source attribution, and Skills CLI installation command.
Verified the published GitHub commit with clean remote Skills CLI and Claude marketplace installs.

Reframed the root README, docs landing page, navigation, and repository reference around a multi-skill collection while preserving the existing HTMA Measure documentation.

Added Zach Prompting to the root skills/ collection.
Reframed the existing backward-compatible htma-measure marketplace plugin as the two-skill ThatGuySam Skills bundle.
Preserved the existing Codex and Claude installation ID while documenting the bundled namespace and the one-skill Skills CLI path.

Added the public Starlight documentation application at apps/docs.
Added installation, usage, prompt, command, troubleshooting, feature, architecture, and reference pages.
Added public llms.txt, llms-full.txt, and llms-small.txt outputs.
Deployed the site at https://skills.samcarlton.com.

Published the standalone htma-measure skill.
Added Codex, Claude Code, and portable plugin manifests.
Added the calibrated memo template, method map, output rubric, and local paid-quote adjustment.
Sanitized personal and workspace-specific markers for public use.