suggest-next-features

Drafts candidate feature entries into a separate suggestions file from git history and progress notes.

754|129|Updated Jun 30, 2026
One-click install
npx skills add https://github.com/Archive228/loopkit --skill suggest-next-features
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: suggest-next-features
Source: https://github.com/Archive228/loopkit/tree/main/skills/suggest-next-features
Command: npx skills add https://github.com/Archive228/loopkit --skill suggest-next-features

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

When every entry in feature_list.json passes or recent sessions stop adding new work, an agent-driven project can stall or silently drift from the spec. This Skill proposes new feature entries without ever mutating the immutable feature_list.json ledger.

Core Features & Use Cases

  • Gap Detection: Compares git commits since the initial scaffold and the last 3 progress notes against feature_list.json to find unlisted behavior and scope creep.
  • Immutable Ledger Protection: Writes 5-10 candidate entries to feature_list.suggestions.json only, never touching feature_list.json directly.
  • Human Merge Workflow: Each suggestion includes a rationale field so an operator can quickly hand-merge chosen entries and discard the rest.
  • Use Case: After all features pass, run this Skill to surface natural next steps like empty states, error handling, or list views implied by shipped endpoints.

Quick Start

Ask the agent to check whether feature_list.json has any remaining failing entries and, if the ledger is exhausted, draft new feature suggestions into feature_list.suggestions.json.

Frequently Asked Questions about suggest-next-features

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I suggest new features without editing feature_list.json?

Write candidate entries to a separate feature_list.suggestions.json file in the same schema, each with passes set to false and a rationale field. A human then hand-merges chosen entries into feature_list.json, preserving its immutability contract.

When should feature suggestions be generated in an agent coding loop?

Generate suggestions only when feature_list.json has zero entries with passes false, or when the last 3 progress entries show no new passes and the user referenced behavior not in the list. A short list alone is not a trigger.

How do I detect scope creep from git history and progress notes?

Run git log since the initial scaffold commit to see what actually shipped, then read the last 3 progress entries in claude-progress.txt. Compare both against feature_list.json to find requested behavior with no matching entry.

Should the feature suggestions file be committed to git?

No. feature_list.suggestions.json is a regenerable proposal, not project state, so it should not be committed and can be added to .gitignore. It is overwritten on each run rather than appended.

What are the limitations of automated feature suggestion?

Suggestions are capped at 10 per run to stay reviewable, and the agent never merges them itself. Entries may duplicate existing scope if feature_list.json is not read in full first, so human review remains mandatory.