agentleFS
hmj1026hmj1026/dhpk/skills/harness-setup/SKILL.md

harness-setup

Skill2 starsChanged 6 days agoMITView on GitHub

Configure or inspect the dhpk harness, or install selected host assets. Not for: ordinary harness audits, application changes, or silent credential/configuration changes. Output: a host-specific setup report with preserved-file, receipt, and terminal evidence.

7 other projects use a version of this.

Adopted by

Most recently taken up by

The file

We split this skill into the rules it gives an agent

3 rules. 0 are common, 0 are distinctive, and 3 are very specific to this project.

  • 0

    Common

    Most of the 4 edited versions we read keep these.

  • 0

    Distinctive

    This author's own choices. Worth a look.

  • 3

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    If --install is present, run the local asset adapter with the explicit source artifact before configuration questions, then report its …

    line 61
  2. Very specific:

    If --show is present, display effective current state and stop without questions or writes.

    line 67
  3. Very specific:

    Verify the resulting state from the host's authoritative settings or receipt.

    line 71

“Common” is measured: kept by at least 6 in 10 of the 4 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Harness Setup

$harness-setup is the explicit setup capability behind the Claude /dhpk:setup front door. It owns setup and configuration evidence; it does not silently broaden a host's authority or invent a second installer.

When to Use

  • The operator explicitly requests $harness-setup or /dhpk:setup.
  • The operator wants to inspect effective setup with --show.
  • The operator explicitly selects a host-supported asset installation group.

When NOT to Use

  • Harness scoring, budget analysis, trimming, or cross-platform synchronization: these retired optional tools are outside the setup contract.
  • Application code, ordinary configuration edits, or credentials not covered by the host setup procedure.
  • A request that supplies a new --host selector. Host is detected from the current runtime; do not add a public host flag.

Host and resource resolution

  1. Detect the current Host from the native invocation context and available tools. A Claude command context selects the Claude adapter; a Codex project-local skill context selects the Codex adapter. If the context is ambiguous, stop BLOCKED with HOST_UNRESOLVED rather than guessing.
  2. Load exactly one package-local adapter reference: Claude setup or Codex setup. A missing reference or declared runtime asset is BLOCKED_RESOURCE_MISSING; do not scan ambient checkouts or remote URLs for a replacement.
  3. Preserve the caller's argument string. The accepted grammar is the argument-hint above; unknown options are BLOCKED. --source-artifact is explicit task input for asset or Codex installation; it is never inferred from a checkout, plugin cache, parent directory, or sibling Skill.

The adapter references, executable wrappers, and procedure resources are Skill-local. scripts/install-assets.sh and scripts/install-codex-project.sh resolve their physical helpers from their own BASH_SOURCE; they execute no adapter selected from the source artifact. The source artifact supplies bytes and manifests only. A missing local helper is BLOCKED_RESOURCE_MISSING; a missing or invalid explicit artifact is SOURCE_ARTIFACT_REQUIRED or SOURCE_ARTIFACT_INVALID.

Workflow

  1. Resolve the consumer project root and read current host state before any write. Existing settings, user files, receipt entries, and unrelated dirty work remain intact.
  2. If --install is present, run the local asset adapter with the explicit source artifact before configuration questions, then report its result and stop. Default --install rules writes .claude/rules/dhpk-overrides.md and does not vendor .claude/dhpk/rules/; pass --vendor only for the discouraged verbatim copy. Do not combine asset installation with interactive reconfiguration.
  3. If --show is present, display effective current state and stop without questions or writes. Otherwise follow the selected adapter's configuration flow, using its native prompt mechanism when available and a visible confirmation boundary before writes.
  4. Verify the resulting state from the host's authoritative settings or receipt. Keep PASS, BLOCKED, UNAVAILABLE, NOT_CONFIGURED, and NOT_RUN distinct; a planned or skipped operation is not a pass.

Output

Return one report with:

Host → operation/arguments → planned/applied/preserved/unsupported items
Evidence → PASS | BLOCKED | UNAVAILABLE | NOT_CONFIGURED | NOT_RUN
Next action → exactly one host-appropriate follow-up

For a write-capable run, include the target root, changed paths, preserved conflicts, receipt or settings evidence, and the command/exit result. Redact keys, tokens, and credentials. Never claim that an installation happened when only a plan or structural receipt was observed.

In the report, remind adopters who use skill handoff saves to add .scratch/ to their .gitignore; do not edit that file.

Verification

  • Host was detected without a new public host flag.
  • Only the selected Skill-local adapter and its declared resources were loaded; artifact files were treated as data.
  • Existing user files and unrelated receipt entries were preserved.
  • --install, --show, and unsupported-option boundaries match the selected Host.
  • The terminal evidence state and one next action are explicit.

References

  • references/claude-setup.md — Claude settings, userConfig, and asset behavior.
  • references/codex-setup.md — Codex project installer, receipt ownership, and Claude-only option boundaries.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.