slash-commanderlisted

# ☄️ Skill: Slash Commander ## 1. How the bridge actually works Real Claude Code discovers slash commands by scanning `.claude/commands/**/*.md` in the host project root — there is no runtime generation step. The bridge is: 1. **Source of truth**: each `workflows/<name>_workflow.md` has a matching, hand-authored `commands/<name>.md` in this submodule (frontmatter `description` + a body that `@`-references the workflow file). 2. **Installation**: `.agents/scripts/install.sh` symlinks `commands/<name>.md` into the host's `.claude/commands/agents/<name>.md`, which Claude Code then exposes as `/agents:<name>`. 3. **Drift check**: `scripts/verify_commands.py` (this skill) scans `commands/*.md` for `@.agents/workflows/...` references and fails if any points at a workflow file that doesn't exist. Run it after adding/renaming a workflow, or as part of `audit_workflow.md`'s `link_audit` step. ## 2. Command Mapping | Workflow | Command | Slash invocation | | :--- | :--- | :--- | | `start_workflow.md` | `commands/start.md` | `/agents:start` | | `close_workflow.md` | `commands/close.md` | `/agents:close` | | `pipeline_workflow.md` | `commands/pipeline.md` | `/agents:pipeline` | | `extract_workflow.md` | `commands/extract.md` | `/agents:extract` | | *(new workflow)* | new `commands/<name>.md` | `/agents:<name>` | ## 3. Governance Constraints - **Isolation**: The installer MUST NOT overwrite anything under host `.claude/` that isn't already a symlink back into `.agents/` — it skips and