Find your next open-source project

Search by name, purpose, or keyword — or paste a GitHub repository URL
Back

michael-denyer/pstack-claude

Installs a set of rigorous agent workflows for planning, reviewing, verifying and shipping code across several coding agents.

  • 68 skills
  • 3.4k stars

The 5 most installed

Ordered by installs on skills.sh
  • unslop1.3k installs

    Removes telltale AI phrasing from any piece of writing.

  • architect1.2k installs

    Sketches types, signatures and module structure before coding, then stays involved as implementation fills in.

  • arena1.2k installs

    Runs several parallel attempts at one task, picks a base, and grafts the best parts of the others into it.

  • Generates a project skill that drives your app the way a real user would to prove it works.

  • how1.2k installs

    Explains how a subsystem works, its runtime flow, and where code should live or which package owns it.

The other 63 skills
  • interrogate1.2k installs

    Runs adversarial multi-model review to break a change or design, keeping only findings proven by a run.

  • reflect1.2k installs

    Reviews the current session for lasting lessons and turns them into edits to existing skills.

  • setup-pstack1.2k installs

    Detects available models and configures which model pstack uses for each role.

  • automate-me1.1k installs

    Turns your preferences and working style into a personal skill that agents follow.

  • babysit1.1k installs

    Watches an open pull request, fixes failing CI, handles simple review comments and drives it toward merge.

  • blast-radius1.1k installs

    Finds what a change could break elsewhere and proves the key assumption by running real code.

  • bro1.1k installs

    Restates the last message in plain everyday language without jargon.

  • deslop1.1k installs

    Removes AI-generated code clutter and tidies up code style.

  • figure-it-out1.1k installs

    Builds an auditable plan for large migrations or ambitious multi-part changes, running a hypothesis loop and logging decisions.

  • Periodically audits a project's verification skill and feature map, driving every feature live and proposing proven corrections.

  • no-comments1.1k installs

    Spawns a comment-checking subagent, fixes accepted findings, and offers ways to encode claimed constraints.

  • Keeps validation and error handling at system edges and business logic in simple pure functions.

  • Has the agent build a reusable script or tool for repetitive work instead of doing it by hand.

  • Turns repeated corrections into automated checks or scripts rather than more written instructions.

  • Builds two or three competing prototypes and compares them before committing to a new design.

  • Guides product and feature tradeoffs toward user delight over implementation convenience.

  • Traces each bug symptom back to where the system first went wrong and fixes it there.

  • Gets core data structures and types right first so the rest of the code becomes straightforward.

  • Sends bulky output to subagents and keeps only summaries in the main conversation.

  • Pushes the agent toward deletion and the smallest change instead of new abstractions and layers.

  • Designs commands and processing steps so reruns after crashes or retries reach the same end state.

  • Moves all callers to a new internal API and deletes the old one in the same change.

  • Simplifies hard-to-follow code by collapsing single-use wrappers and shrinking mutable scope.

  • Helps the agent model the domain properly when writing stateful logic or heavily branching code.

  • Lets the agent proceed on reversible work instead of stopping to ask permission.

  • Keeps planned rewrites and migrations aimed at the target design without throwaway compatibility code.

  • Verifies against the real artifact before declaring a task done, not just that it compiles.

  • Rethinks the design from first principles when adding a new requirement to existing code.

  • Removes shared state between concurrent actors before resorting to locking or serialization.

  • Removes dead code and redundant pieces first, then builds the new work on the simpler base.

  • Designs types so illegal states can't be represented and external data is parsed at boundaries.

  • recall1.1k installs

    Rebuilds a brief of your recent work from chat history and records so you can resume.

  • show-me-your-work1.1k installs

    Keeps a row-per-decision log so long or unattended runs can be reviewed later.

  • swarm1.1k installs

    Runs many parallel workers on a task and returns one combined report.

  • technical-writing1.1k installs

    Applies a layered technical-writing standard to docs, readmes, PR descriptions and commit messages.

  • Runs a very strict maintainability review focused on abstractions, giant files and tangled conditions.

  • Applies TypeScript best practices whenever the agent reads or edits .ts or .tsx files.

  • why1.1k installs

    Explains why a design or behavior exists, with cited decisions, tradeoffs and thresholds.

  • fix-ci1k installs

    Finds failing pull request checks, inspects logs or external links, and applies focused fixes.

  • Resolves merge conflicts without interaction, validates the build and tests, and finalizes the resolution.

  • get-pr-comments1k installs

    Fetches and summarizes review comments from the active pull request.

  • Prepares a pull request for review by cleaning history, improving descriptions and adding reviewer guidance without changing behavior.

  • Breaks multi-step work into small units that each end in a verifiable state.

  • teach998 installs

    Explains a change or subsystem plainly by combining how it works with why it was built.

  • Summarizes the commits you authored over a chosen time period into a short update.

  • Writes tests that assert observable behavior so they survive internal refactors.

  • Stops the agent from repeating failed fixes and instead tests the underlying assumption directly.

  • tdd935 installs

    Writes a failing test first, then the fix, when you ask for test-driven work.

  • Checks what limits a measured number and rules out misleading measurements before trusting it.

  • correct611 installs

    Turns repeated agent mistakes in a repo into checks that make them impossible, proving each check catches a real past error.

  • Checks a performance measurement for sound limits, repeatability and relevance before you report or act on it.

  • poteto-help576 installs

    Guides you through setup and helps pick the right skill, playbook or principle for a task.

  • ketchup428 installs

    Catches you up on what changed in the chat and lists every item waiting on your decision.

  • thermos416 installs

    Runs the bug/security and code-quality reviews in parallel, then merges their findings.

  • Audits a branch's changes for bugs, breaking changes, security issues and feature-gate leaks.

  • sharpen380 installs

    Deepens a task rather than expanding it, then rewrites your prompt accordingly.

  • in-your-own-words363 installs

    Restates what it thinks your goals are and the problem you are trying to solve.

  • Generates a project skill that formally verifies risky code such as concurrency, state machines or parsers.

  • Updates a project's formal verification after code changes, reruns it, and says whether failures are drift or bugs.

  • statechart207 installs

    Checks a TLA+ model of a stateful system and draws it as a Mermaid statechart.

  • ketchup-video198 installs

    Catches you up with a short subtitled slide video of what happened and the pending decisions.

  • pstack-routing1 installs

    Points the agent to the right pstack workflow before starting each task.

  • Makes the agent reply concisely, write clean prose and code, and verify its work in a consistent style.