Problem
The writing-plans skill hardcodes two execution options in its "Execution Handoff" section:
- Subagent-Driven (recommended)
- Inline Execution
The plan header template also bakes this in:
For agentic workers: REQUIRED SUB-SKILL: Use superpowers:subagent-driven-development (recommended) or superpowers:executing-plans
Users who define custom execution modes in CLAUDE.md (e.g., agent teams with Builder/Inspector/Director roles) get overridden by the skill's prescriptive flow — even though superpowers explicitly says CLAUDE.md takes precedence over skill defaults.
What happens
- User's CLAUDE.md has a triage table mapping task types to execution modes (conversation, subagent, agent teams, etc.)
writing-plans finishes a plan and presents only subagent-driven or inline- The model follows the skill's prescriptive options instead of checking the user's triage table
- User has to manually correct every time
Suggestion
The execution handoff could check for user-defined execution modes in CLAUDE.md/AGENTS.md before presenting options. If custom modes exist, include them. If not, fall back to the current two defaults.
This would respect the priority order superpowers already documents: user instructions > skill defaults.
Problem
The
writing-plansskill hardcodes two execution options in its "Execution Handoff" section:The plan header template also bakes this in:
Users who define custom execution modes in CLAUDE.md (e.g., agent teams with Builder/Inspector/Director roles) get overridden by the skill's prescriptive flow — even though superpowers explicitly says CLAUDE.md takes precedence over skill defaults.
What happens
writing-plansfinishes a plan and presents only subagent-driven or inlineSuggestion
The execution handoff could check for user-defined execution modes in CLAUDE.md/AGENTS.md before presenting options. If custom modes exist, include them. If not, fall back to the current two defaults.
This would respect the priority order superpowers already documents: user instructions > skill defaults.