Shared agent setup and reusable skill definitions.
This repo is intended to be the source of truth for reusable, cross-project skills. Harness-specific wrappers (Codex, Claude, etc.) should be generated from skills-core/ rather than hand-maintained in each project.
skills-core/: Canonical, harness-neutral skills.statusline/: Reusable Codex and Claude status-line config.scripts/sync-skills-all.sh: Sync/render script for target repos.scripts/install-statusline.sh: Install status-line config on a machine.adapters/: Adapter notes and conventions.
Run from this repo:
cd /home/ag9898/projects/ag.dev
./scripts/sync-skills-all.sh --target /path/to/target-repo --symlink-codexWhat it does by default:
- Syncs all skills under
skills-core/*(dynamic; not hardcoded by name) - Renders
.claude/skills/*with{{CMD_PREFIX}}=/ - Renders
.agents/skills/*with{{CMD_PREFIX}}=$ - With
--symlink-codex, links.codex/skills/*to.agents/skills/*
Useful options:
./scripts/sync-skills-all.sh --help
./scripts/sync-skills-all.sh --target /path/to/repo --dry-run
./scripts/sync-skills-claude.sh --target /path/to/repo
./scripts/sync-skills-codex.sh --target /path/to/repo --symlink-codexRun from this repo:
cd /home/ag9898/projects/ag.dev
./scripts/install-statusline.shThis merges the Codex [tui] status-line settings into ~/.codex/config.toml,
copies the Claude status-line command to ~/.claude/statusline-command.sh, and
sets statusLine in ~/.claude/settings.json.
After syncing into a target repo, invoke skills in the harness chat:
- Slash-style examples:
/query-workboard,/start-task - Dollar-style examples:
$query-workboard,$start-task
Loop example:
/ralphloop start-task iterations:3$ralphloop start-task iterations:3