gh:work-x

Work-X Execution Command

Execute work with the normal shipping discipline, plus Morph-X safeguards for iOS Swift/ObjC output. This reduces template-code repetition risk; it does not guarantee App Review success and does not replace real product, UI, content, metadata, or feature differentiation.

Input Document

<input_document> #$ARGUMENTS </input_document>

Execution Workflow

Phase 0: Input Triage

Determine how to proceed from <input_document>.

Plan document: read it completely in Phase 1 and treat it as a decision artifact.

Bare prompt: scan likely files, nearby tests, and local conventions; classify scope:

Complexity	Signals	Action
Trivial	1-2 files, no behavioral change	Setup, apply directly, and run relevant tests if behavior changed
Small / Medium	Clear scope, under ~10 files	Build a task list, then execute
Large	Cross-cutting, architecture, auth/payments/migrations, 10+ files	Recommend planning/brainstorming; if proceeding, build tasks carefully

Phase 0.1: Task Lifecycle Start

Run:

gale-task log skill_started --skill gh:work-x --title "${ARGUMENTS:-work-x}" 2>/dev/null || true

If gale-task is unavailable, skip silently. This must never block execution.

Phase 0.6: HKTMemory Retrieve

Before choosing a blueprint, retrieve related implementation context and historical Morph-X fingerprints:

hkt-memory retrieve \
  --query "<task summary, iOS components, Swift/ObjC files, prior blueprint or strategy tags>" \
  --layer all --limit 10 --min-similarity 0.35 \
  --vector-weight 0.7 --bm25-weight 0.3

Use results only as context: extract constraints, avoided patterns, and prior blueprint/strategy fingerprints. Do not copy historical code shape. If no results or any command error occurs, proceed silently without blocking.

Phase 0.6b: HKTMemory Session Search

Query related work sessions:

hkt-memory session-search \
  --query "gh:work-x <task title or iOS feature summary>" \
  --limit 5

Use returned sessions as supplementary context for blueprint selection. If unavailable, continue silently; session search is non-blocking.

Phase 1: Quick Start

1. Read Plan and Clarify when a plan/spec path was provided.

   • Read the full document, including requirements, scope boundaries, implementation units, files, test scenarios, verification, and deferred unknowns.
   • Ask only when ambiguity blocks correct execution.
   • Get user approval to proceed when the source workflow requires it.

2. Setup Environment

   • Check the current branch and default branch.
   • If already on a meaningful feature branch, continue unless the user asks otherwise.
   • If on the default branch, create a feature branch or worktree unless the user explicitly confirms working on the default branch.
   • Respect workspace safety: do not overwrite uncommitted user work.

3. Create Task List

   • Derive tasks from implementation units, dependencies, file ownership, tests, and verification.
   • Preserve plan U-IDs in task names.
   • Include quality checks and relevant tests.

4. Choose Execution Strategy

   • Use inline execution for small work or interactive tasks.
   • Use serial subagents for dependent multi-unit work.
   • Use parallel subagents only after checking file ownership overlap; do not stage, commit, or run the full suite inside parallel workers.

Phase 1.5: Morph-X Blueprint Constraints

Before writing or changing Swift/ObjC code:

1. Load .morph-config.yaml from the target iOS project root if present. If missing, use in-memory defaults and record config_source: default.
2. Build a blueprint prompt block from: project seed, feature intent, retrieved historical fingerprints, blacklisted pattern tags, existing architecture, and local test/build constraints.
3. Select a blueprint that differs from recently used strategy fingerprints across:
   • state management
   • module boundaries
   • file splitting
   • protocol or abstraction boundaries
   • error handling
   • data-flow organization
4. Write the constraints into the task notes before implementation. The constraints should guide original structure; they must not request cosmetic churn or semantic changes.
5. If config, CLI, or memory retrieval is unavailable, continue with a manual blueprint summary and mark Morph-X as degraded, not failed.

Phase 2: Execute

For each task:

1. Mark it in progress.
2. Read relevant files and local patterns.
3. Find adjacent tests before editing.
4. Implement with existing conventions and the Phase 1.5 blueprint constraints.
5. Run a System-Wide Test Check: if shared contracts or user-facing behavior changed, choose a relevant broader test target; otherwise run focused tests.
6. Assess quality before moving on: behavior covered, edge cases considered, no unrelated refactors, no user work overwritten.

Phase 3: Morph-X Apply and Similarity Audit

After Swift/ObjC code is produced and before final summary:

1. Apply safe transformation when the CLI is available:

   gale-harness morph --apply --config .morph-config.yaml --report .morph-report.json
   

   The transform must be semantics-preserving. If the command, SwiftSyntax, or ObjC adapter is unavailable, skip the transform, record the reason, and continue to audit fallback.

2. Audit similarity:

   gale-harness audit --similarity --config .morph-config.yaml --report .morph-audit.json
   

   Include AST/structure fingerprint, token n-gram, statement Jaccard, and control-flow approximation results when available. Missing baselines or tools are non-blocking and should be reported as degraded signals.

3. Handle thresholds:

   • Default behavior is warn, not block.
   • If .morph-config.yaml configures a blocking threshold and the audit exceeds it, stop before finalizing code and ask the user how to proceed.
   • Never claim the audit proves App Review compliance.

4. Run focused compile/test verification appropriate to the changed iOS project when available. If no Xcode or Swift test target is runnable, state the gap.

Phase 4.5: HKTMemory Store

After implementation and Morph-X audit, store a concise memory record:

hkt-memory store \
  --content "<summary with repo-relative files, blueprint constraints, strategy fingerprint, audit status, degraded fallback notes>" \
  --title "<work-x title>" \
  --topic "work-x morph blueprint strategy fingerprint" \
  --layer all

Store only summaries, tags, and fingerprints; do not store full source code. On error, note it as non-blocking and do not fail the workflow.

Phase 5: Close

Return:

## Work-X Summary
**Completed**: [what changed]
**Blueprint**: [constraints and strategy fingerprint]
**Morph Apply**: [report path or degraded reason]
**Similarity Audit**: [report path, threshold result, blocking/warning decision]
**Tests**: [commands run and results]
**Compliance Boundary**: Reduces template-code repetition risk; does not guarantee App Review success.

Run gale-task log skill_completed 2>/dev/null || true. If unavailable, skip silently.