Build Plan — Slider Autonomy (FEAT-011)

Source identity (from 04-build/implementation-plan.md):

ddx:
  id: helix.implementation-plan
  depends_on:
    - TP-002
    - TD-011
  status: superseded
  superseded_by: helix.prd

SUPERSEDED — This build plan was scoped to slider autonomy (FEAT-011) and helix run --autonomy, which are no longer part of HELIX’s scope per the current PRD (helix.prd). This document is retained for historical context. New implementation planning for HELIX must focus on artifact schema completion, template metadata, the portable alignment skill, and public documentation projection — not CLI execution-loop features.

Build Plan — Slider Autonomy (FEAT-011)

Scope

This build plan scopes the implementation of FEAT-011 (slider autonomy): the operator-facing autonomy slider that lets helix run start at a chosen activity (input, frame, design, build, review) instead of always beginning at intake. The plan covers wrapper changes, default-path proof, and public-doc updates so the shipped surface matches the governing design.

Governing Artifacts:

docs/helix/01-frame/features/FEAT-011-slider-autonomy.md — feature spec
docs/helix/02-design/technical-designs/TD-011-slider-autonomy-implementation.md — technical design
docs/helix/03-test/test-plans/TP-002-helix-cli.md — wrapper-CLI test plan
docs/helix/02-design/contracts/CONTRACT-001-ddx-helix-boundary.md — DDx/HELIX boundary

Shared Constraints

Tracker-first execution: every story-level slice is a ddx bead under .ddx/beads.jsonl. No build work begins without a claimed bead.
helix run --autonomy <activity> must be backward-compatible: the default remains input for operators who run helix run with no flag.
The CLI surface is a compatibility wrapper over DDx-managed execution per CONTRACT-001. Build issues add behavior to HELIX prompts and policy, not to a parallel execution stack.
Deterministic proof gates every change: bash tests/helix-cli.sh and bash tests/validate-skills.sh must pass before any build bead closes.
spec-id on every build bead points at the nearest governing artifact (FEAT-011, TD-011, or TP-002).

Build Sequencing

Order	Story / Area	Governing Artifacts	Depends On	Notes
1	Wrapper accepts `--autonomy <activity>` and dispatches	FEAT-011, TD-011 §3	None	Smallest landable slice; unlocks downstream proof work
2	`helix input` skill end-to-end against new wrapper	FEAT-011, TD-011 §4	1	Intake path is the most-used autonomy stop; prove it before opening lower-stop work
3	Lower-stop autonomy entrypoints (`frame`, `design`, `build`)	FEAT-011, TD-011 §5	1	Reuse the dispatch layer from #1
4	Public-doc and demo refresh	FEAT-011, FEAT-007	1, 2, 3	Update website intake examples and demo recordings only after the surface stabilizes
5	Deterministic proof for every autonomy entrypoint	TP-002, TD-011 §6	1, 2, 3	Extend `tests/helix-cli.sh` with one positive + one rejection case per stop
6	Cross-model review wiring on autonomy-driven runs	TD-011 §7, TP-002	5	Verify `--review-agent` still works when the run starts mid-activity

Issue Decomposition

Story-level work is tracked via ddx bead in .ddx/beads.jsonl.

Per-issue requirements:

Labels: helix, activity:build, kind:build, area:cli, plus story:FEAT-011-<slice> for grouping.
References: FEAT-011, TD-011, TP-002, and this build plan.
spec-id pointing at the nearest governing artifact.
Blockers expressed as dependencies[] on the predecessor bead.

Story / Area	Goal	Dependencies
Wrapper dispatch (`helix run --autonomy`)	`helix run --autonomy <activity>` selects the right skill entry and rejects unknown activities	none
`helix input` against new wrapper	Intake path drives `ddx agent execute-loop` with the right governing-artifact context	wrapper dispatch
Lower-stop entrypoints	`frame`, `design`, `build` autonomy stops route to their corresponding skill prompts	wrapper dispatch
Public docs and demos	Website intake docs + asciinema recordings reflect the slider surface	all wrapper work
Proof harness	`tests/helix-cli.sh` covers one happy + one rejection case per stop, plus one cross-model review test	wrapper + intake + lower-stop work
Cross-model review wiring	`helix run --autonomy <activity> --review-agent <other>` still produces a deterministic review pass	proof harness

Quality Gates

Failing tests exist in tests/helix-cli.sh before each wrapper change is implemented
All required tests pass before closing a build issue (bash tests/helix-cli.sh, bash tests/validate-skills.sh)
Behavior changes update FEAT-011 / TD-011 / TP-002 explicitly
helix review is run after each landed slice; review findings file new beads rather than being absorbed silently
No build bead closes without a spec-id and a passing review

Risks

Risk	Impact	Response
Lower-stop autonomy exposes ungoverned skills	High	Refuse `--autonomy <activity>` when the corresponding skill prompt is missing or stale; surface as `BLOCKED` rather than running
Intake-path proof drifts as default behavior changes	Medium	Lock the intake test cases to the published `helix input` contract; treat changes as governing-doc evolutions, not test updates
Operators bypass the slider via direct `ddx agent execute-loop` calls	Low	Document the slider as the supported entry; the direct DDx path remains available but unsupported as a HELIX surface
Cross-model review wiring regresses when the run starts mid-activity	Medium	Add a cross-model review test for at least one mid-activity autonomy stop

Exit Criteria

Build issue set is defined with sequence and dependencies (above table is current)
Shared constraints are documented and traced to FEAT-011, TD-011, and CONTRACT-001
Verification expectations are explicit: tests/helix-cli.sh, tests/validate-skills.sh, and a cross-model review pass on at least one autonomy stop
FEAT-011 acceptance criteria are mapped to specific bead IDs in the queue

Review Checklist

Governing artifacts are listed and exist on disk
Shared constraints trace back to requirements (FEAT-011), design (TD-011), and contract (CONTRACT-001)
Build sequence has a justified ordering (wrapper before consumers; proof before public-doc updates)
Dependencies between build steps are explicit
Each story/area references its governing artifacts (FEAT-011, TD-011, TP-002)
Issue decomposition follows tracker conventions (labels, spec-id, deps)
Quality gates are specific and enforceable (named test scripts, named artifacts)
Risks have concrete responses, not vague strategies
Plan is consistent with TP-002 and TD-011