jmuncor's personal collection of Claude Code skills.

• senior-planning-loop — converge a plan/design doc to implementation-readiness: fan out independent agents across diverse lenses, refute every finding and decision by default, re-ground every claim, and loop in rounds until dry — producing a CONVERGED doc with D-* decision records and FM-* failure modes.

agent-skills/
├── .claude-plugin/
│   ├── plugin.json        # plugin manifest (this repo = one plugin)
│   └── marketplace.json   # makes the repo installable via /plugin marketplace add
├── skills/                # FLAT — one kebab-case folder per skill, each with a SKILL.md
├── template/SKILL.md      # copy this to start a new skill
├── scripts/install.sh     # symlink skills/* into ~/.claude/skills/ for local use
├── CHANGELOG.md
└── LICENSE

Categories live here in the README, not as nested folders — keep skills/ flat.

cp -r template skills/my-new-skill      # kebab-case folder name = the /command name
$EDITOR skills/my-new-skill/SKILL.md    # fill in frontmatter + steps
./scripts/install.sh                    # symlink it into ~/.claude/skills/

The name/description in the frontmatter is what Claude matches on to decide when to fire the skill — write description in third person, lead with the use case, and include the words you'd naturally say. Keep SKILL.md lean and push detail into sibling files (reference.md, scripts/) that load only when needed.

./scripts/install.sh

Symlinks each skill into ~/.claude/skills/. The repo stays the source of truth, and edits to a SKILL.md hot-reload in Claude Code. Run --dry-run to preview.

/plugin marketplace add jmuncor/agent-skills
/plugin install agent-skills@agent-skills

MIT