local-skills
v0.1.0Reference skill bodies for the `local` cartridge — every role bound to the filesystem, no tracker anywhere.
0 GitHub starsUpdated 1 hour ago
Directory evidence
- Runtimes
- Claude Code
- Parsed components
- 23 skill or MCP entries
- Source updated
- Sep 25, 2026
- Manifest status
- Canonical path parsed
The directory validates manifest shape and source location. It does not execute the plugin or provide a security endorsement. Review the indexing methodology →
Install local-skills for Claude Code
claude plugin marketplace add IchenDEV/agent-plugin-mkt
claude plugin marketplace update agent-plugin-marketplace
claude plugin install local-skills@agent-plugin-marketplacePaste and run these commands in a terminal with Claude Code. They add and refresh the PluginsMP catalog, then install this plugin.
The installer fetches third-party code from the source repository shown on this page. This directory validates manifest structure and source location, but does not perform a security audit; review the manifest, components, and source before installing.
Get the source manually
git clone https://github.com/ppfenning/coxswain-cartridgesClone the source repository, then follow its setup instructions to add the plugin to a compatible client. The plugin root is skills-plugins/local-skills/.
Plugin files
├── .claude-plugin/plugin.json├── skills/arbitrate/SKILL.md├── skills/arbitrate-plans/SKILL.md├── skills/attack-plan/SKILL.md├── skills/build-in-worktree/SKILL.md├── skills/decompose-initiative/SKILL.md├── skills/dispatch-graphs/SKILL.md├── skills/handoff-brief/SKILL.md├── skills/plan-task/SKILL.md├── skills/reconcile-state/SKILL.md├── skills/release-sync/SKILL.md├── skills/retro-runs/SKILL.md├── skills/review-adversary/SKILL.md├── skills/review-charter/SKILL.md├── skills/route-work/SKILL.md├── skills/scope-work/SKILL.md├── skills/style-pass/SKILL.md├── skills/triage-classify/SKILL.md├── skills/update-docs/SKILL.md├── skills/validate-chunk/SKILL.md├── skills/validate-phase/SKILL.md├── skills/verify-evidence/SKILL.md├── skills/workstore-item/SKILL.md└── skills/workstore-state/SKILL.md
Included Skills23
Resolve reviewer disagreement with a decision and its price — side with an argument, not a role.
Choose between two independent plans for one work item, or merge them into a better one — and name what the choice costs.
The adversary of a plan — attack the claims it rests on before a build budget is spent on it.
Execute a plan and return the change as a unified diff. The worktree is yours; nothing else is.
Turn one large idea into phases and a task DAG whose edges are real — parallelism falls out of honest dependencies.
Given the registry, the work store, the intake queue, ledger stats, and the runs already in flight, pick which graphs run next and why.
Check that what one step produced is what the next step actually needs — stop the line when the artifact is missing, send it back once when only its evidence is.
Turn one work item into an ordered, bounded plan another agent can execute without asking questions.
Given declared state and actual state, name each drift and the smallest correction that closes it.
Decide whether the umbrella's CLI docs, manifest, release notes, and package pages match what the components actually ship, and report each mismatch as a Drift.
Read the run record and say what should change — proposals with citations, never vibes.
The second reviewer, whose job is to disagree — attack the claims the change rests on.
Review a change against the team's own written standards — the charter decides, not your taste.
Decide whether a request changes a repository and so must be routed, or is answered inline — and if it routes, size it, launch it, and report on it without touching the work yourself.
Decide what shape one piece of work is — epic, parent with subtasks, or a single item — before anyone plans it.
Make the narrow style edit an approved diff still needs, or return nothing — this seat has no verdict to give.
Match one alert to a runbook entry by symptom — fast, honest about confidence, and never inventive.
Apply arm for the knowledge base — land one approved doc correction, in the document's own voice.
Decide whether ONE task actually satisfied its own description — from evidence, never from the builder's word for it.
Decide whether these tasks, TOGETHER, accomplish the phase's original goal — independent of whoever owns the phase.
Run the deterministic checks that make a claim true or false — and say which, with the output in hand.
Apply arm for work items — create or update exactly one markdown work item, exactly as approved.
Apply arm and SINGLE WRITER for work state — move one item to one state, and nothing else.
Plugin manifests1
{
"$schema": "https://anthropic.com/claude-code/plugin.schema.json",
"name": "local-skills",
"version": "0.1.0",
"description": "Reference skill bodies for the `local` cartridge — every role bound to the filesystem, no tracker anywhere."
}For maintainers
If you maintain this plugin, link to this source-backed listing from your README so users can review its manifest and indexed components.
[local-skills on Agent Plugins Marketplace](https://pluginsmp.com/plugins/local-skills)