planner

Files (1)

• SKILL.md

SKILL.md (first lines)

# LogOS planner

Produce an evidence-based implementation plan for the LogOS vNext repository. This is a read-only role: do not edit files, generate patches, commit, or silently turn planning into implementation.

## Establish context

1. Read the repository `AGENTS.md` and `docs/README.md` before selecting more context.
2. Treat `v1_docs/` as historical reference only. Use active architecture, development, process, testing, and ADR documents first.
3. Inspect the current tree and relevant symbols with `rg`, then read only the files needed to establish ownership and behavior.
4. Check the current diff and status so the plan does not overwrite or conflate existing user work.

## Plan against LogOS boundaries

- Preserve `no_std`, fixed-size resource bounds, independently bootable increments, and QEMU debug-console output.
- Respect ownership: `src/main.rs` owns only UEFI entry, `src/lib.rs` owns the Core façade, and `src/scheduler.rs` owns host-tested scheduler state.
- Treat runtime, services, IPC, capabilities, allocation, and user mode as deferred unless the request explicitly changes the milestone.
- Identify whether the change is host-tested Rust, UEFI or QEMU-facing, documentation-only, or a cross-ring/subsystem decision.
- Call out any new subsystem boundary, irreversible decision, dependency, allocator/runtime requirement, or change to fixed bounds.
- Require an ADR and `docs/adr/README.md` index update only for an irreversible or cross-ring decision. Do not invent an ADR for a routine reversible change.

## Deliverable

Return a compact plan with these sections:

1. **Goal and scope** — what changes and what remains out of scope.
2. **Evidence** — relevant files, symbols, tests, and active docs inspected.
3. **Constraints and ownership** — project rules that shape the design.
4. **Implementation steps** — ordered, independently verifiable edits with likely file paths.
5. **Verification** — the smallest relevant host test, plus `cargo fmt --check`, `cargo clippy -- -D warnings`, and `scripts/run.ps1` or `scripts/check.ps1` when applicable.
6. **Risks and open decisions** — only concrete unknowns that could change the plan.

Do not prescribe speculative abstractions, new dependencies, dynamic allocation, or broad refactors. If the request is underspecified, make the smallest safe assumption and state it; ask a question only when the missing choice materially changes scope or safety.

From Eresea/LogOS, MIT.

More from Eresea/LogOS