Help us improve
Share bugs, ideas, or general feedback.
From helix
Insight-driven planning. Recalls memory, explores informed by insights, decomposes via planner, writes plan.
npx claudepluginhub enzokro/crinzo-plugins --plugin helixHow this skill is triggered — by the user, by Claude, or both
Slash command
/helix:helix-meta-plannerThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
Plan-mode-only skill. Produces an implementation plan informed by helix's accumulated project insights.
Replicates Claude Code plan mode in read-only fashion: explores codebase, designs changes, gets user approval, then writes plan to .claude/plans/<slug>.md without executing. Triggers on 'plan only' phrases.
Generates detailed work plans using agent teams: Prometheus for user interviews and requirements, Metis for gap analysis, saved to .sisyphus/plans/. For complex refactors, builds, or architecture tasks.
Conducts interview-driven planning with 20-40 clarifying questions, 5-15 parallel Explore agents for codebase analysis, verbatim requirements capture, and plan validation. For new features, refactoring, architecture changes, migrations.
Share bugs, ideas, or general feedback.
Plan-mode-only skill. Produces an implementation plan informed by helix's accumulated project insights.
HELIX="$(cat .helix/plugin_root)"
python3 "$HELIX/lib/injection.py" strategic-recall "$ARGUMENTS"
Parse JSON. Use summary for triage, synthesize insights into blocks:
_effectiveness >= 0.70): decomposition rules, verification needs, sequencing._effectiveness < 0.40) or derived/failure tags: flag for extra verification, smaller tasks._hop: 1 insights (graph-adjacent, not direct match). Treat as exploration targets.Triage signals: coverage_ratio > 0.3 = well-mapped, trust constraints. < 0.1 = uncharted, expand exploration. graph_expanded_count > 0 = graph surfacing related context.
If recall returns empty -- proceed without constraints; first sessions have no memory.
Map areas relevant to both objective and insight-identified targets.
git ls-files | head -80 -- identify 3-6 natural partitions.subagent_type="helix:helix-explorer", model=sonnet, max_turns=30. All in ONE message. Prompt: SCOPE: {partition}\nFOCUS: {focus}\nOBJECTIVE: $ARGUMENTS.Obvious scope (single module, clear file set): skip swarm, use Glob/Grep/Read directly.
Spawn planner: subagent_type="helix:helix-planner", max_turns=500. Prompt: OBJECTIVE: $ARGUMENTS\nEXPLORATION: {merged_findings_json}\nCONSTRAINTS: {constraints_from_recall}\nRISK_AREAS: {risk_areas_from_recall}. Omit empty blocks. Parse PLAN_SPEC JSON array.
If decomposition raises questions -- use AskUserQuestion to resolve before synthesis.
Write the plan file (path from system context):
# {Objective summary}
## Context
Why this change is needed — the problem, what prompted it, intended outcome.
## Insights Applied
Relevant helix insights and how each shaped the plan:
- [eff%] insight content → influenced {which decision}
## Key Files
Files identified by exploration, grouped by concern:
- {area}: `file1.py`, `file2.py` — {what they do, why they matter}
## Implementation Plan
### 1. {slug} (seq)
{description}
- **Files:** relevant_files
- **Depends on:** blocked_by (or "none — parallel")
- **Verify:** command
### 2. {slug} (seq)
...
## Verification
How to test the complete change end-to-end.
Quality bar: A developer reading the plan should know exactly what changes, in what order, verified how — without re-exploring the codebase.
Call ExitPlanMode to present the plan for user approval.