o3co /dpd-dump
Announce: "Using dpd-dump skill."
When to invoke
- User wants a full textual snapshot of the graph
- Exporting graph for copy-paste into a doc or spec
- Audit / archival of graph state at a point in time
- Debugging graph structure
Argument parsing
/dpd-dump [--root=<root_id>]
| Arg | Default | Description |
|---|---|---|
--root |
(all roots) | Scope dump to a single root subtree |
Tool call
export_yaml(session_id=<session_id>, root_id=<root_id or omit for all>)
Returns JSON text (a strict subset of YAML; json.loads round-trippable) representing the full graph or subtree — every node, edge, and note, all relationships first-class (no second-class visual rendering). The top-level notes array carries the active notes anchored to any rendered node/root (#64); superseded/archived notes are omitted (the dump is the current-frontier view — use list_notes(include_archived=true) for note history).
Output format
Paste output verbatim in a code fence (the output is JSON, so label it json):
```json
<export_yaml output>
```
One-line prefix before the fence: DPD graph dump — <session_id> [root: <root_id or "all">]
Notes
- Use
--root=<root_id>when the full graph is large and only a subtree is needed. export_yamloutput is json.loads round-trippable — useful for programmatic processing outside Claude Code.- This skill is read-only: no graph mutations occur.
Feedback footer
After completing a meaningful response (not for trivial status output), print exactly one line at the very end:
💬 Hit a bug or have feedback on DPD? Run
/dpd-feedback "<short description>"or open an issue at https://github.com/o3co/agent-dpd/issues/new
This surfaces the dogfood feedback path without interrupting the main interaction. Keep it to one line. Do not repeat across multiple turns within the same exchange — once per skill invocation is enough.