skills.sh
AvailableUniversal skill installation, including Claude Code and Codex.
npx skills add https://github.com/ulpi-io/skills-autonomous-engineering --skill auto-planauto-plan
Compiles a testable spec into a validated DAG of atomic, independently verifiable build slices.
WORKFLOW / 05/auto-plan <spec path, or the feature to plan (will look for .ulpi/spec/*)>
- 01
Read the spec and verify its repository assumptions.
- 02
Decompose it into coherent tasks with at most three scoped paths each.
- 03
Wire real output dependencies and produce topological execution layers.
- 04
Run the deterministic structural validator.
- 05
Adversarially review coverage, paths, independence, and graph correctness until clean.
▸ operating envelope
Use it deliberately.
The shortest useful definition of where this skill helps—and where it should stay out of the way.
Use when
- A grounded specification needs an implementable task breakdown.
- Several changes can run in parallel once their dependencies are explicit.
- The build needs exact file scopes, validation commands, and specialist routing.
Do not use when
- The requested behavior is still vague or untestable; return to auto-spec.
- You need to decide what should be built rather than how to build it.
- You are asking the agent to implement the plan; use auto-build.
▸ output contract
What must be true when it stops.
Produces
- .ulpi/plans/<name>.json
- Dependency edges and topological execution layers
- Acceptance-criterion-to-task coverage map
- Specialist assignments and self-review result
Guarantees
- Every task maps to real paths or paths created by an earlier dependency.
- Tasks are atomic, independently verifiable, and carry slice-scoped validation.
- The graph is acyclic and execution layers are topologically valid.
- Write overlap is ordered rather than raced.
- A stalled review reports remaining defects and never signs off the graph.
Honest failure states
- invalid — the deterministic graph validator rejects the plan
- stalled — adversarial review leaves unresolved plan defects
- ungrounded — a task depends on phantom paths, commands, or requirements
▸ enforcement
Know which claims are executable.
Scripts are linked when this skill owns deterministic machinery. The remaining rules are explicit operating contracts and must not be marketed as hooks.
▸ install and invoke
One skill, three honest surfaces.
Claude Code plugin
AvailableInstall the full plugin, then invoke /auto-plan or let routing select it from context.
Codex plugin
Work in progressA work in progress on the codex-native-plugin branch — build it, start a new session, then invoke $autonomous-engineering:auto-plan.
▸ composes with