skills.sh
AvailableUniversal skill installation, including Claude Code and Codex.
npx skills add https://github.com/ulpi-io/skills-autonomous-engineering --skill auto-mapauto-map
Builds a verified, disclosure-tiered context architecture that loads repository knowledge only when needed.
WORKFLOW / 05/auto-map [scope — full refresh (default), a package/dir, or 'verify' to audit the existing map]
- 01
Detect project shape and inventory existing context files and ownership.
- 02
Discover real directories, commands, boundaries, and conventions.
- 03
Write the smallest always-on root, path-scoped rules, nested maps, and linked deep references.
- 04
Run the anti-lie validator and execute claimed commands.
- 05
Report budgets, preserved human content, drift, and unverified claims.
▸ 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 project lacks a trustworthy context map.
- Architecture-level changes moved packages, commands, or boundaries.
- An existing map needs a drift audit or the pipeline has just shipped meaningful work.
Do not use when
- A routine edit did not change the project architecture.
- The goal is to invent coding policy rather than map observed reality.
- The user did not request this durable project-memory mutation and no invoked workflow composes it.
▸ output contract
What must be true when it stops.
Produces
- Lean root CLAUDE.md
- Path-scoped .claude/rules/*.md files
- Nested CLAUDE.md files for significant directories
- Verified command and drift report with per-tier line counts
Guarantees
- Paths, exports, routes, and commands are checked against the repository.
- Root CLAUDE.md stays at or below 150 lines, rules at 120, and nested maps at 100.
- Deep references are linked for on-demand reading rather than imported at startup.
- Generated sections update without clobbering human-written instructions.
- Commands that cannot run are labeled unverified instead of asserted true.
Honest failure states
- drift — a mapped fact no longer matches repository reality
- unverified — a claimed command could not be executed
- over_budget — content belongs in a deeper disclosure tier
- unsafe_clobber — existing human-owned content cannot be preserved confidently
▸ 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-map 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-map.
▸ composes with