Skip to content

Compatibility

  • Tease: The collection installs through the documented paths, with evidence kept separate from model-quality claims.
  • Lede: Each canonical skill uses the standard skills/<name>/SKILL.md layout, and the shared marketplace bundle exposes the same source directories.
  • Why it matters: Installation success proves packaging and discovery; it does not prove that every future output will be correct or better.
  • Go deeper: Review the dated checks, then keep the compatibility boundaries attached to any downstream claim.

Verified on 2026-07-11 against public commit 6989b11 or a later documentation-only commit:

SurfaceResultEvidence
Open Skills CLI 1.5.14PassedFound exactly one skill, copied all six files, and produced a source/hash lock record.
Codex plugin 0.1.1PassedMarketplace add, plugin add, installed/enabled listing, and cached skill files all succeeded from a clean CODEX_HOME.
Claude Code plugin 0.1.1PassedMarketplace add and plugin install both succeeded from a clean CLAUDE_CONFIG_DIR.
Manual copyCompatible by formatThe skill is a self-contained directory with relative resource paths.

The Open Skills CLI’s documented discovery locations include the root skills/ directory. Its CLI reference supports repository shorthand, --list, --skill, agent targeting, global installs, updates, and removal.

Claude’s marketplace documentation supports same-repository relative plugin sources. Version 0.1.1 uses that pattern to avoid a second GitHub SSH clone.

Verified on 2026-07-15 against the local publication candidate:

SurfaceResultEvidence
Open Skills CLI 1.5.14PassedDiscovered two skills, selected only zach-prompting, copied its three files into a clean temporary Codex target, and produced a source/hash lock record. The installed directory matched the canonical source.
Claude marketplace bundlePassedStrict manifest validation, local marketplace add, and clean plugin install all succeeded. The stable htma-measure install key loaded both canonical SKILL.md files without differences.
JSON and YAML metadataPassedAll distribution JSON parsed; agents/openai.yaml parsed; the official skill validator reported Skill is valid!.
Docs and machine corpusPassedStarlight built 26 pages and all three llms*.txt outputs; the docs-spec gate passed 11 checks with no warnings or failures.

Post-push verification also passed at skill publication commit 54ac5d4: GitHub main matched the commit, the Skills CLI installed only zach-prompting from the remote repository with no content differences, and a fresh Claude marketplace installation resolved version 54ac5d427ee9 with both skill directories intact.

Verified on 2026-07-15 against public commit 8405b24:

SurfaceResultEvidence
Remote marketplace installPassedA fresh Claude configuration added the public repository over HTTPS and installed the stable key htma-measure@thatguysam-skills.
Collection namespacePassedThe installed .claude-plugin/plugin.json reported name: sam and skills: ./skills.
Component inventoryPassedclaude plugin details identified the plugin as sam and listed exactly htma-measure and zach-prompting.
Installed contentPassedRecursive comparison found no differences between either cached skill directory and the public commit’s canonical source.
Manifest validationPassedclaude plugin validate . --strict accepted the marketplace and plugin structure.

Claude’s official plugin marketplace documentation distinguishes the stable marketplace entry used by install commands from the plugin name used to namespace bundled skills. A fresh interactive invocation was not run during this packaging check; the verified manifest and component inventory are the basis for the /sam:<skill> compatibility claim.

  • Each published skill works independently of the other published skills; HTMA Measure may still use optional companion skills when they are available.
  • Clients must support Agent Skills-style Markdown discovery or allow the instructions to be loaded manually.
  • A successful install proves packaging and discovery, not the quality of every future estimate or prompt revision.
  • Current-source research still requires network access and appropriate browsing tools.
  • Private inputs remain the user’s responsibility. HTMA Measure does not fetch or infer inaccessible financial facts.