feature-deep-dive
Build a code-verified mental model of one feature or subsystem. Use before owning, debugging, changing, reviewing, or redesigning a complex area that needs a technical trace from entry point to data, jobs, providers, failures, and tests.
- License 许可
- MIT
- 最近更新
- 2026-09-20
- Stars
- ★ 4
- Forks
- 0
01 前言 Front Matter
- name
- feature-deep-dive
- description
- Build a code-verified mental model of one feature or subsystem. Use before owning, debugging, changing, reviewing, or redesigning a complex area that needs a technical trace from entry point to data, jobs, providers, failures, and tests.
03 SKILL.md 正文
Feature Deep Dive
Build debug-grade understanding of one feature or subsystem. Do not change code unless the user asks.
One responsibility: make a developer understand one area well enough to debug/change it without guessing.
Scope
Choose one vertical slice or subsystem only: invite claim, promotion scheduling, content generation callback flow, search ranking, billing checkout, etc.
If the user needs the whole-system map first, use rebuild-mental-model. If this trace changes the global model, update or hand off to rebuild-mental-model.
Output
Follow repo conventions first. If none exist, create/update:
docs/feature-deep-dives/<feature-slug>.mdCONTEXT.mdif domain language changes and it exists; otherwisedocs/ubiquitous-language.md
Never include secrets or env values. Provider/env var names are fine.
Workflow
- Frame - user/business goal, trigger, success condition, core object/state, in/out boundaries.
- Trace - entry point, API/service/helper path, persistence/source of truth, jobs/events, providers/webhooks, permissions, tests.
- Pressure-test - state transitions, idempotency, partial failure, retries, races, legacy behavior, observability, first debug handles.
- Update docs - write the feature doc, update ubiquitous language when terms drift, record global architecture impacts.
- Teach/verify - explain layer by layer, ask for restatement when useful, correct gaps before calling the area understood.
Feature Doc Contract
- Verification coverage
- Overview
- Actor and boundary map
- Runtime path
- State transitions
- Source of truth
- Providers and external effects
- Permissions and security notes
- Failure modes, retries, and idempotency
- Observability/debug playbook
- Test surface
- Global architecture impact
- Open questions and follow-ups
- Quick recall
Verification Labels
Mark claims explicitly:
verified- confirmed from code/config/schema/tests.partial- traced but not end-to-end.not verified- plausible or doc-derived only.drift- docs and code disagree.
Never present intended future behavior as current code.
Quick Recall Contract
End every feature doc with:
## Quick Recall
- What this feature does:
- Main trigger:
- Source of truth:
- Top files/boundaries:
- Top tables/models:
- Providers:
- First debug checks:
- Biggest trap:
- Global architecture impact:
Completion Standard
Before finishing:
- One feature doc is updated.
- Verification labels are present.
- Ubiquitous language is updated or explicitly unchanged.
- Debug handles and tests are named.
- Global architecture impacts are listed, even when the answer is
None.
04 同仓库更多 Skills
writing-shape
Take a markdown file of raw material and shape it into an article through a conversational session — drafting candidate openings, growing the piece paragraph by paragraph, arguing about format (lists, tables, callouts, quotes) at each step. Use when the user has a pile of notes, fragments, or a rough draft and wants help turning it into something publishable.
writing-fragments
Grilling session that mines the user for fragments — heterogeneous nuggets of writing (claims, vignettes, sharp sentences, half-thoughts) — and appends them to a single document as raw material for a future article. Use when the user wants to develop ideas before imposing structure, or mentions "fragments", "ideate", or "raw material" for writing.
writing-beats
Shape an article as a journey of beats, choose-your-own-adventure style. The user picks a starting beat from the raw material, you write only that beat, then offer options for where to pivot next, beat by beat, until the article reaches a natural end. Use when the user has raw material and wants to assemble it as a narrative rather than an argument.
stop-slop
Remove AI writing patterns from prose. Use when drafting, editing, or reviewing text to eliminate predictable AI tells.
humanizer
Remove signs of AI-generated writing from text. Use when editing or reviewing text to make it sound more natural and human-written. Based on Wikipedia's comprehensive "Signs of AI writing" guide. Detects and fixes patterns including: inflated symbolism, promotional language, superficial -ing analyses, vague attributions, em dash overuse, rule of three, AI vocabulary words, passive voice, negative parallelisms, and filler phrases.
feedback
Prepare, deliver, and debrief timely interpersonal feedback using a question-first coaching flow, observable behavior, owned experience, and concrete impact. Use when drafting feedback, coaching managers, rehearsing a feedback conversation, praising specific behavior, or reviewing whether feedback landed.