About

Some prompts snippets I have found useful and [mostly] effective through my agentic coding workflows.

General-use

/research

  • 1 Read {X{,Y,Z)}} in depth, understand how it works deeply, what it does and all its specificities.
  • 1 Study {X{,Y,Z)}} in great details, understand the intricacies of it and
    • write a detailed report of your learnings and findings in research.md
    • help me refactor…
    • help me test…
    • help me debug…
    • help me implement…
  • Please ask me any clarifying and follow-up questions.

/plan

  • 2 Make the plan extremely concise. Sacrifice grammar for the sake of concision.
  • 2 At the end of each plan, give me a list of unresolved questions to answer, if any.

/tdd

  • BDD → TDD.

/implement

  • 3 “take your time, don’t stop till X” (x=done or condition)

/refactor

  • What should we refactor next?
  • Where is complexity increasing?
  • Don’t care about backwards compatibility.

DRY

  • Can you identify DRY opportunities between X and Y implementations?
  • Can you identify significant simplification opportunities to reduce cognitive load?
  • Can you identify separation of concerns?

Design Patterns

  • Any opportunities to better follow design patterns such as, but not limited to:
    • Domain-driven design (DDD)
    • RESTful API design
    • Repository patterns / query objects / data gateways and adapters
    • Service objects/classes
    • Consistently typed contracts
    • Error-handling consistency

/execute

  • Let’s consider {X} in a phased order, so we can produce semantic, atomic, and clear commits.

/test


  • What tests are missing?

  • Are there any missing important tests when compared to similar implementations X/Y/Z?

  • Are there any re-organization opportunities to homogenize & reduce cognitive load (compared to other implementations)?

/debug

  • 1 Go through {X{,Y,Z)}}, understand it deeply and look for potential bugs. There definitely are bugs in the system as it sometimes {DO A(,B,C)}. Keep researching the flow until you find all the bugs, don’t stop until all the bugs are found. When you’re done, write a detailed report of your findings in research.md

/docs

  • Help me write a concise and summarized yet comprehensive PR description.
    • Backend developers should be able to take a quick glimpse at the implementation details.
    • If API contracts are introduced or modified, include a section for Frontend developers to quickly review new or updated contracts and relevant client-side concerns.
    • Focus on the latest changes for the current Git branch, unless indicated differently.

/agents (meta)

  • Can we incorporate these new helpers into relevant agent artifacts?
  • Revisit:
    • AGENTS.md
    • .agents artifacts

Coder vs Reviewer roles

OK, I went ahead and made the refactor/fix myself. Can you compare it against your recommendations/plan and identify constructive critiques, gaps, or improvements I might be missing?


Trivelta/PAM-specific

/migrate-to-v2

Read current implementation {v1} in depth. Understand how it works, what it does, its specificities, and its intricacies.

When that is done, make a phased plan to migrate this surface area completely to @services/admin_panel/admin_panel_v2/ under the most appropriate context, with these technical guidelines:

  • Follow all v2 conventions and design patterns strictly.
  • Find opportunities to:
    • Decrease complexity and reduce cognitive load.
    • Identify separation-of-concerns and DRY opportunities.
  • On existing versus new functionality:
    • Prioritize backward compatibility for consumers (Frontend) on existing functionality.
      • Apply TDD initially by replicating/translating existing tests into the v2 context first.
    • Add new features (such as {…, …, …}, and potentially others) on top of current functionality at the end of the implementation.

As an optional follow-up phase, completely remove the dead v1 surface.

Make the plan extremely concise. Sacrifice grammar for concision.

Ask me any unresolved, clarifying, or follow-up questions and decisions, if any.

Output should be a series of /gh-stack PRs that are reviewable and concise, with /pr-description and atomic, semantic commits.


Citations

Footnotes

  1. boristane.com/blog/how-i-use-claude-code 2 3

  2. aihero.dev/my-agents-md-file-for-building-plans-you-actually-read 2

  3. https://x.com/steipete/status/2040542898773762231?s=46&t=ssP6EM3-WGWpWUN96NHzSA