Index of every page on this site: llms.txt
All skills
Run it yourself

Use when adopting existing harness configuration into Loadout, setting up Loadout in a project or global source, or changing Claude, Codex, Droid, OpenCode or Pi configuration. Handles onboarding, permissions, instructions, skills, agents (subagents), settings, hooks, plugins and MCP through their authoritative sources.

  • 4 files
  • 52 KB
  • Updated 5 days ago
  • GitHub
Use this Skill: https://skilld.dev/gh/nielsmadan/agentic-coding/loadout

Nothing lands on disk. Nothing to clean up.

Fork this Skill

Edit a local copy. It keeps the author and licence.

SKILL.md

≈83 tokens for metadata: the name and description. ≈1.3k when used: this file. ≈12k more on demand in 3 files.

Description uses 4.1% of example budget

Before choosing a Skill, your Agent reads its name and description. All available Skills share that space.

  • A shorter description leaves more room for other Skills. This entry exceeds our 1% size suggestion.

In our Claude Code example, all Skill names and descriptions share 8,000 characters. This Skill uses ≈331 characters, or 4.1%.

The 1% threshold is a size suggestion. Longer descriptions can still fit.

Your model, settings, and other Skills decide how much text your Agent can read.

Example settings and source

The example uses a 200k-token context and default Claude Code settings. The count includes the name, description, separators, and when_to_use when present. Codex also counts local file paths.

Skit's source and limits: Codex 0.160.1, Claude Code 2.1.292.

Loadout

Edit the declared source for each configured consumer. Generated harness files are outputs.

Instructions

  1. For setup or migration, read onboarding. Preview with loadout init project --dry-run --json or loadout init global --dry-run --json, resolve harnesses and ownership, then apply the approved plan. Never execute discovered scripts or installers.
  2. For ordinary changes, read configuration routing. Default to personal configuration for the current project; explicit project/global scope takes precedence. Locate the existing config and follow its declared producers: the manifest's harness blocks and the source slices they name. Ask before widening scope or changing a source shared by other consumers. A repository that has not adopted loadout, where the user wants personal configuration, is covered by a stealth source if one targets it: edit the stealth source's files and sync with loadout sync --stealth <name>, never init project or a new loadout/ there; see the stealth row of the scope table and “Stealth sources” in the same reference.
  3. Apply a generic request to configured harnesses with a sound mapping. Report unsupported consumers. Keep MCP server definitions separate from MCP tool-approval policy. Preserve ordering, comments, private sources and unrelated fields; never create overlapping producers.
  4. Sync the edited source: loadout sync --root <repo> for project/personal changes; loadout sync --global for the active global profile, or loadout sync --global --profile <name> when explicitly selected; loadout sync --stealth <name> for a stealth source. Rendered files are output: edit the source folder. Ask whether an unnamed change under a non-default profile belongs to that profile or its inherited default.
  5. Report source files changed, harnesses reached, unsupported consumers and the sync result.

Examples

  • “Set up Loadout here” previews discovery, resolves the scope and configured harnesses, and shows writes and removals before applying. Success leaves editable, uncommitted source.
  • “Allow just test for me” edits the existing personal permissions producer and syncs. If the project has no personal producer, ask before changing committed ownership.
  • “Add my own MCP server for the work repositories” in a repository a stealth source targets edits the stealth source's mcp.toml and runs loadout sync --stealth <name>; it does not touch the repository's .mcp.json or opencode.json.
  • “Change the global Pi model” follows the settings fragment Pi's harness block names, edits its defaultModel value and syncs the selected profile. Pi renders settings only when its block also declares plugins.

Troubleshooting

  • An unresolved preview needs an explicit scope, a harness choice, or a fix to the file it names; --yes only approves a resolved plan. Retain unsupported inputs and report their paths.
  • A missing personal producer requires a scope decision. Do not silently edit shared source or invent a second contributor claiming the same keys.
  • A change made outside loadout belongs in the source, not beside it. When sync refuses an output, or check reports hand-edited, owned-key, owned-key-removed, unknown, unmanaged, unmanaged-link or unparsable, follow “Porting a change made outside loadout” in configuration routing. Run sync --force only after showing the diff and getting the user's approval; when sync cut a diff short, read the complete one in the sync-abort.diff file it names. A file holding a secret is never diffed: loadout names it and counts the lines that differ. Read the file on disk and the source directly to see the change, or move the secret to an environment variable.
  • A project.instructions or <harness>.instructions note that the document repeats a heading, or a kit add line naming a project skill the kit already supplies, marks guidance held twice. Resolve it by “Guidance held twice” in configuration routing.
  • An init that lists unadopted items left copies that disagree across harnesses in place. Merge them by “Resolving conflicting copies at init” in onboarding.
  • An init that lists secrets adopted them into the source as they were. Say the source's remote must stay private and offer to move each into an environment variable, by “Secrets in the adopted source” in onboarding.
  • A failed init has already put everything back. An init killed part way stops every later init until it is put back, which needs the user at a terminal, as the onboarding reference describes.
  • In a linked git worktree of the registered global source, the checkout and merge hooks print that nothing was synced: a worktree is not the registered folder, so it never renders global scope. Edit and commit there as usual (its pre-commit check still runs), then sync from the registered source once the change reaches it.

Source: SKILL.md on GitHub

No rule matched.

skilld matched fixed text patterns in SKILL.md and file names. Patterns miss obfuscated code.

skilld run checks every file with the same patterns. It asks for approval before it loads a Skill with a behavior marked Needs approval.

No third-party reports yet.

Signed by skilld at 9fbe49f. This ties the file your Agent reads to that commit on GitHub. It does not review the instructions.

Last checked against GitHub 4 days ago.

Activeupdated 5 days ago

README badge

README badge for nielsmadan/agentic-coding/loadout