agentsclimarketplace

Harness init

Skill atlax-tech/harness-armor/skills/harness-init

Self-improving, tool-first Harness Engineering Agent Skills for Claude Code, Codex, Cursor and TRAE. Build, audit, update and verify AGENTS.md-based repositories.

Install
npx -y skills add atlax-tech/harness-armor --skill harness-init

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

3 things to look at

  • 26 days oldThe repository was created 26 days ago. New is not bad, but a brand new repository carrying a familiar-sounding name is the shape a typosquat arrives in, and there has been no time for anyone else to find a problem with it.
  • no licenseNo license file was found in the repository. Code published without one is not open source by default, so using it at work is a question for whoever answers licensing questions where you are.
  • 1 stars1 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

What its author says it does

Copied from the file, not written here

Create a fact-safe Harness Engineering skeleton for an empty or basic-files-only repository, including a short AGENTS.md, project document stubs, and managed state. Use when starting an undefined repository or explicitly invoking harness-init. Do not use when substantive product documents or business code already exist, and never guess the product or technology stack.

The file declares its own license as CC-BY-NC-4.0. That is the author’s claim about this one file, and it is not the same thing as the license GitHub reports for the repository, which is listed with the other numbers below.

SKILL.md

4.4 KB, as published. Nobody here has run it

Initialize an undefined repository

Create a useful knowledge skeleton without filling unknowns with plausible fiction.

Inputs

  • Repository root.
  • Optional project name or known purpose supplied by the user.
  • Existing README, LICENSE, .gitignore, and editor files are preservation constraints, not blank space.

Load the contract

  1. Treat the directory containing this file as SKILL_ROOT.
  2. Read references/init-policy.md.
  3. Resolve shared from SKILL_ROOT/../../shared or, for direct installs, SKILL_ROOT/../.harness-armor.
  4. Read spec/harness-engineering-v1.md and the templates under templates/. Stop with an installation error if the shared root is absent.

Applicability gate

  1. Read applicable AGENTS.md instructions if any exist.
  2. Run python scripts/detect_repository_state.py <root>.
  3. Continue only for EMPTY, including repositories containing only basic administrative files.
  4. If the state is DOCS_ONLY, stop and route to harness-build. For code, route to harness-promotion. For any Harness or conflict, route to harness-check.

Fact ledger

Before writing, list every supplied fact and every required unknown. Use CONFIRMED only for user statements or existing file evidence. Mark everything else UNRESOLVED using the exact marker in assets/unresolved-marker.md. Do not infer users, features, language, framework, database, deployment, or commands.

Write workflow

  1. Propose the new-file set. The default set is:
    • AGENTS.md;
    • docs/PRODUCT.md, ARCHITECTURE.md, DESIGN.md, DEVELOPMENT.md, TESTING.md, ACCEPTANCE.md, and ROADMAP.md;
    • .harness/manifest.json, source-index.json, and unresolved.json.
  2. Treat an explicit request to initialize as authorization to create only missing files in that set. If any target exists, do not overwrite it; show the collision and request a narrower decision.
  3. Create a short AGENTS.md knowledge map from the shared template.
  4. Create focused document skeletons. Include only confirmed facts and explicit unresolved markers. Do not create empty filler sections merely to increase coverage.
  5. Create the state files with ownership for files actually created. Record pre-existing files as user or observed, never managed.
  6. Create docs/decisions/README.md and docs/development-log/README.md only when the client cannot preserve empty directories and those maps add useful instructions.
  7. Run python scripts/validate_harness_structure.py <root> and relevant local reference checks.
  8. If validation fails, remove only newly created files from this run when safe; otherwise report the partial state exactly. Never alter pre-existing files during cleanup.

Allowed changes

Create missing Harness files and directories listed in the approved plan. Do not modify existing repository files, business code, package metadata, CI, or Git configuration.

Forbidden behavior

  • Do not create an implementation, dependency manifest, framework scaffold, or pretend build/test command.
  • Do not rewrite README or LICENSE.
  • Do not mark unresolved placeholders CONFIRMED.
  • Do not silently replace an earlier partial Harness.

Idempotency and failure handling

A second run on an unchanged initialized repository must produce no file diff; route it to harness-check. Report collisions, permission failures, incomplete runtime, truncated scans, and unrun validators. Never claim initialization succeeded when structure validation did not complete.

Result format

Report: applicability evidence, confirmed facts, unresolved items, files created, files preserved, validation commands with actual results, and manual acceptance steps. End by recommending that the user fill docs/PRODUCT.md and then invoke harness-build once substantive product sources exist.

Gives 0 of the 12 instructions most project setup skills give

Counted across 999 of the 1,637 authors here whose files we hold, read 2026-08-06

  • ask one question at a timein 29 of 999, across 28 files
  • detect the package manager from lockfilesin 28 of 999, across 9 files
  • present findings to the userin 25 of 999, across 4 files
  • explore current repo statein 24 of 999, across 3 files
  • update the agent skills block in place if it existsin 24 of 999, across 3 files
  • install husky lint-staged and prettierin 23 of 999, across 4 files
  • create the lintstagedrc filein 22 of 999, across 3 files
  • commit all changed filesin 22 of 999, across 3 files
  • run lint-staged to verify it worksin 22 of 999, across 3 files
  • initialize huskyin 21 of 999, across 2 files
  • create the husky pre-commit filein 21 of 999, across 2 files
  • create a prettierrc file if missingin 21 of 999, across 2 files

Said here and by no other author read

  • create missing knowledge skeleton files
  • read the init policy and shared spec
  • run the repository state detection script
  • list all confirmed facts and unresolved unknowns
  • mark non-confirmed items as unresolved
  • create short knowledge map from template

Grouped from the skills themselves: near-identical wordings counted once, and counted by distinct author, so one author publishing three of these counts once.

Keep looking

Skills are one crate of 328,083. Ordering is by how many stacks a row turns up in, so the top of any crate is what has actually been picked rather than what has the most stars.