omp-skills

A personal skills and command library for Oh My Pi. The content targets OMP's built-in tool surface and execution workflows.

Installation

OMP discovers user skills one directory below ~/.omp/agent/skills/ and command files in ~/.omp/agent/commands/. Clone the repository, then symlink each entry:

git clone https://github.com/fgladisch/omp-skills.git ~/.omp/omp-skills
mkdir -p ~/.omp/agent/skills ~/.omp/agent/commands
for skill in "$HOME"/.omp/omp-skills/skills/*; do
  ln -sfn "$skill" "$HOME/.omp/agent/skills/$(basename "$skill")"
done
for command in "$HOME"/.omp/omp-skills/commands/*.md; do
  ln -sfn "$command" "$HOME/.omp/agent/commands/$(basename "$command")"
done

For local development, replace "$HOME"/.omp/omp-skills with the path to this checkout. If upgrading from code-review, remove its old symlink from ~/.omp/agent/skills/ and rerun the installation loops to discover local-code-review and the restored pr-create command. Restart OMP after adding or removing skills or commands so discovery runs again.

Commands

Command	Purpose
local-review-cycle	Review local changes and apply fixes until LGTM
pr-create	Create a PR and address review feedback until LGTM

Skills

Skill	When to use
commit	When the user asks to commit changes; follows repository conventions with gitmoji as the fallback
finalizing-changes	When a coherent set of changes is ready to validate and then commit, merge, publish as a PR, or retain
local-code-review	When reviewing local code changes in a working tree, branch, commit, or defined file set
planning	While plan mode is active to shape plan contents; also for plan-only requests or consequential unresolved decisions
systematic-debugging	When a bug, failing test, build failure, performance regression, or unexpected behavior needs diagnosis
test-driven-development	For behavior changes and bug fixes when a test or executable regression check can establish the expected result first
verification-before-completion	Before claiming work is complete, fixed, passing, reviewed, ready, committed, or integrated

Skill dependency graph

flowchart TB
  P["planning<br/>(when requested or decisions remain)"]
  SD["systematic-debugging"]
  TDD["test-driven-development"]
  CR["local-code-review"]
  V["verification-before-completion"]
  C["commit"]
  F["finalizing-changes"]

  SD -. "when a regression cycle is feasible" .-> TDD
  TDD -. "when change risk warrants review" .-> CR
  CR --> V
  TDD --> V
  SD --> V
  F -. "when the outcome requires a commit" .-> C
  C --> V
  F --> V

Loading

Arrows mean the source skill explicitly applies the target skill, and dashed edges are conditional. Planning is an independent entry point; clear implementation requests can proceed without it. TDD requests independent review only when compatibility, security, data-integrity, or operational risk warrants the additional pass.

OMP tooling

• ask provides structured decisions; no extension-provided selection tool is required.
• read, glob, grep, lsp, edit, and write handle repository inspection and changes before shell fallbacks.
• debug and browser verify runtime and UI behavior directly.
• task, hub, and todo provide parallel agents, process supervision, and phase tracking.
• Skills and their supporting files are addressable through skill://<name>.
• Project instructions live in AGENTS.md; OMP loads them as repository context.

License

MIT — see LICENSE.