overseer-plan
Convert markdown planning documents to Overseer tasks via MCP codemode. Use when converting plans, specs, or design docs to trackable task hierarchies.
How do I install this agent skill?
npx skills add https://github.com/dmmulroy/overseer --skill overseer-planIs this agent skill safe to install?
- Gen Agent Trust Hubpass
The skill converts markdown planning documents into trackable tasks by executing JavaScript through an MCP API. It includes a defensive 'Oracle Review' step to validate task breakdowns. The primary risk is indirect prompt injection from the markdown files being processed, which could lead to incorrect task creation or manipulation.
- Socketpass
No alerts
- Snykpass
Risk: LOW · No issues
- Runlayerwarn
2/4 files flagged
What does this agent skill do?
Converting Markdown Documents to Overseer Tasks
Use /overseer-plan to convert any markdown planning document into trackable Overseer tasks.
When to Use
- After completing a plan in plan mode
- Converting specs/design docs to implementation tasks
- Creating tasks from roadmap or milestone documents
Usage
/overseer-plan <markdown-file-path>
/overseer-plan <file> --priority 1 # Set priority (0-2, 0=highest)
/overseer-plan <file> --parent <task-id> # Create as child of existing task
What It Does
- Reads markdown file
- Extracts title from first
#heading (strips "Plan: " prefix) - Creates Overseer milestone (or child task if
--parentprovided) - Analyzes structure for child task breakdown
- Creates child tasks (depth 1) or subtasks (depth 2) when appropriate
- Returns task ID and breakdown summary
Hierarchy Levels
| Depth | Name | Example |
|---|---|---|
| 0 | Milestone | "Add user authentication system" |
| 1 | Task | "Implement JWT middleware" |
| 2 | Subtask | "Add token verification function" |
Breakdown Decision
Create subtasks when:
- 3-7 clearly separable work items
- Implementation across multiple files/components
- Clear sequential dependencies
Keep single milestone when:
- 1-2 steps only
- Work items tightly coupled
- Plan is exploratory/investigative
Task Quality Criteria
Every task must be:
- Atomic: Single committable unit of work
- Validated: Has tests OR explicit acceptance criteria in context ("Done when: ...")
- Clear: Technical, specific, imperative verb
Every milestone must:
- Demoable: Produces runnable/testable increment
- Builds on prior: Can depend on previous milestone's output
Review Workflow
- Analyze document → propose breakdown
- Invoke Oracle to review breakdown and suggest improvements
- Incorporate feedback
- Create in Overseer (persists to SQLite via MCP)
After Creating
await tasks.get("<id>"); // TaskWithContext (full context + learnings)
await tasks.list({ parentId: "<id>" }); // Task[] (children without context chain)
await tasks.start("<id>"); // Task (VCS required - creates bookmark, records start commit)
await tasks.complete("<id>", { result: "...", learnings: [...] }); // Task (VCS required - commits, bubbles learnings)
VCS Required: start and complete require jj or git (fail with NotARepository if none found). CRUD operations work without VCS.
Note: Priority must be 1-5. Blockers cannot be ancestors or descendants.
When NOT to Use
- Document incomplete or exploratory
- Content not actionable
- No meaningful planning content
Reading Order
| Task | File |
|---|---|
| Understanding API | @file references/api.md |
| Agent implementation | @file references/implementation.md |
| See examples | @file references/examples.md |
In This Reference
| File | Purpose |
|---|---|
references/api.md | Overseer MCP codemode API types/methods |
references/implementation.md | Step-by-step execution instructions for agent |
references/examples.md | Complete worked examples |
How can the creator link this skill?
Add the canonical catalog link to the repository README so users can inspect current installs and available audits. The publishing guide covers the complete discovery path.
<a href="https://skillzs.dev/skills/dmmulroy/overseer/overseer-plan">View overseer-plan on skillZs</a>