# Skills

Weyaw installs project-local skills for coding agents. Initialize the project,
then select the integration target:

```bash
aw init "Short project summary"
aw install --target agents
```

Use `--target claude-code` for Claude Code, `--target codex` for the legacy
Codex compatibility directory, or `--target all` for every supported target.

## Common skills

- `$aw` — load lightweight project workflow context without starting a task.
- `$aw-init` — initialize or reconcile the project profile.
- `$aw-task <intention>` — create or locate one roadmap-linked task and run its
  lifecycle.
- `$aw-resume <task-id>` — continue the recorded next step.
- `$aw-status` — summarize task lifecycle and blockers.
- `$aw-health` — diagnose or repair workflow integrity.
- `$aw-help` — suggest a project-aware next action.
- `$aw-blueprint` — update durable design language and standards.
- `$aw-roadmap` — maintain roadmap phases and items without creating tasks.
- `$aw-index` — inspect index capability, freshness, and discovery surfaces.
- `$aw-commit` — stage and commit one archived task in project and workflow
  repositories.
- `$aw-goal <item ids>` — run a bounded ordered batch, committing after each
  archived task.

## Static and dynamic guidance

Installed `SKILL.md` files define stable orchestration rules. At activation,
skills load dynamic guidance from the current project, including core context
from `.aw/blueprint/aw.md`. One centralized guidance block also provides the
absolute path to the version-matched product docs shipped with the installed
package; individual skill files do not repeat that path. Dynamic guidance can narrow evidence, render matching official Agent triggers, or describe runtime boundaries; it does not replace direct source reads. `complex.reasoning` may activate in any Weyaw context, while project-writing AGT-X is limited to official Brainstorm, Analyze, and Execute triggers with frozen-worktree, common-sandbox, and explicit write-root readiness evidence. `$aw-task` and `$aw-resume` may also render `control.aw-cli-chain` to open one temporary broad `aw **` capability under a device-local runtime owner; the harness cannot edit workflow files directly and the Host rereads resulting state.

Use the packaged `aw` command in installed projects. Within the Weyaw source
checkout, use `node bin/aw-debug.js` so guidance reflects the current code.

## Host-owned step todos

The `$aw-task` Host seeds and updates durable ordered todos for `analyze`,
`plan`, `verify`, and `review` through
`aw step todo set <task-id> <step> --stdin`. The command atomically replaces
the complete list, permits at most one `doing` row, and requires every retained
row to be `done` before the tracked step can complete. The Host appends newly
discovered work before performing it and explicitly removes obsolete rows by
replacement; generated skills and Pi do not own these mutations.

`execute` uses `plan.json` plus `execution.json` instead of task-status todos.
Its final ordered list takes Plan-backed titles from `plan.json`, overlays their
statuses from `execution.json`, then appends execution-only rows using their own
titles. Plan-backed execution rows therefore omit `title`. After execution
begins, newly discovered work is appended to `execution.json` as a titled
execution-only row with explicit deviation evidence before the work runs. The
approved plan is not silently rewritten.

## Agent ownership and authority

Task Agent runs keep durable task-linked session evidence. Generic non-task reasoning and control operations use device-local runtime-operation owners instead of creating placeholder tasks. Project config can enable or disable official definitions, choose worker order, and narrow bounded turns, fan-out, and content budgets; extensions may change recognized official behavior prose but cannot create triggers or widen context, writable, or root authority.

Prompt contracts own decomposition, ordered execution, retry behavior, and evidence reporting. Core hard gates own official trigger/context matching, task or operation identity, writable agreement, frozen worktree selection, canonical write-root containment, readiness, and resource conflicts. The Host remains the sole owner of workflow gates, artifact semantics, acceptance, lifecycle, and commits.

See [Concepts](concepts.md) and [Artifacts](artifacts.md).