Repository navigation
fix(skills/planning-a-feature): stop the writing-plans handoff question - #94
Merged
Merged
Conversation
Step 6 loads `superpowers:writing-plans`, and that skill ends with an "Execution Handoff" section. The section tells the agent to ask the user for an execution method, subagent-driven or native. Nothing in `planning-a-feature` said that the section does not apply. So an agent could ask that question after the plan and the state file were saved, and not invoke `feature-dev-workflow:developing-a-feature` as step 9 requires. Step 6 now says that the sub-skill ends at its self-review, and that step 9 replaces its handoff. A new Red flags row names the question and points to steps 7, 8, and 9. In a baseline on the unedited skill, agents wrote a real plan and state file in a scratch repository. The two runs on the smallest model both asked the execution-method question. Three runs on each of the two larger models did not. With the new text, five runs on the smallest model asked only the setup questions of `developing-a-feature`. Fixes #93 Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Contributor
There was a problem hiding this comment.
🟡 Changes recommended
The unconditional prohibition conflicts with a required single-PR implementation path and should be explicitly scoped to the planning handoff.
1 open finding
What changed in this PR
Clarifies the planning workflow’s handoff to implementation.
Changes:
- Stops
writing-plansfrom asking about execution methods. - Adds a matching red-flag entry.
| File | Description |
|---|---|
skills/planning-a-feature/SKILL.md |
Clarifies handoff ownership between planning and development. |
🧠 Review effort: Balanced
💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.
…the plan Step 6 said, with no limit, not to invoke `superpowers:subagent-driven-development` or `superpowers:executing-plans`. Step 4 of `developing-a-feature` requires `superpowers:subagent-driven-development` for a single PR with 2 or more independent tasks. An agent could carry the rule into that step and skip the required skill. Step 6 now names no execution skill. It says not to start execution from the "Execution Handoff" section, and that `developing-a-feature` decides which execution skill it uses. With the new text, five runs on the smallest model still asked only the setup questions of `developing-a-feature`. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Owner
Author
|
Review round 1, for the review of 128a02f.
|
This was referenced Oct 12, 2026
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Description
Fixes #93
After
planning-a-featuresaved the plan and the state file, an agent could ask the user to choose "subagent-driven" or "native" execution, and not invokefeature-dev-workflow:developing-a-feature. That question is the "Execution Handoff" section ofsuperpowers:writing-plans, which step 6 loads, and nothing told the agent that the section does not apply. Step 6 now says that the sub-skill ends at its self-review and that step 9 replaces its handoff. The plan-review question of that same section ("does the plan capture what you want?") is not ruled out, and some agents still ask it next to the setup questions.Changes
superpowers:writing-plans, to ask no execution-method question when the plan is saved, and not to start execution from the "Execution Handoff" section. It says thatfeature-dev-workflow:developing-a-featuredecides which execution skill it uses.## Red flagshas a new row that names the "subagent-driven or native?" question and points to steps 7, 8, and 9.Testing
Each test run was a fresh agent in its own scratch repository with an approved spec on a feature branch. The agent read
planning-a-feature, loadedsuperpowers:writing-planswith the Skill tool, wrote a real plan and state file, committed and pushed them, and ended its turn. A run failed when its last message asked the user for an execution method.On the unedited skill, the two runs on the smallest model both failed, and one said that it had not started
developing-a-feature. Three runs on each of the two larger models passed, so the gap did not show there. On the edited skill, five runs on the smallest model passed: each asked only the setup questions ofdeveloping-a-feature. After a review finding changed the text of step 6, five more runs on the smallest model passed in the same way. The edited skill was not run again on the two larger models, and no run used a single-PR feature.The change touches no script and no JSON file, so no script test applies.
🤖 Generated with Claude Code