cursor

pstack

Bundle of 14 skills

Use it when:: You want the agent to go deep before it ships: understand the code, settle the design, then prove the change works.
It does:: Runs multi-model workflows that explain, design, attack and check changes, plus writing rules and short engineering principles the workflows cite.

how, why, architect, arena, interrogate, blast-radius, technical-writing, unslop, principle-boundary-discipline, principle-type-system-discipline, principle-fix-root-causes, principle-prove-it-works, principle-guard-the-context-window, principle-encode-lessons-in-structure

Age: 5 months
Stars: 15.8K; #71
Total Downloads: 33.6K; top 21%
Weekly Downloads: 2.7K; top 18%

unslop

Part of pstack

Use it when:: A draft, doc or message reads as machine-written and you want it plain.
It does:: Scans the text for 28 numbered patterns, then rewrites it, keeping the meaning and intended tone.

command none none anti-slop writing

Age: 5 months
Stars: 15.8K; #71
Total Downloads: 33.6K; top 21%
Weekly Downloads: 2.7K; top 18%

thermo-nuclear-code-quality-review

Use it when:: You want a harsh review of a branch's code quality, not its bugs: abstractions, sprawling files, ad-hoc conditionals, needless wrappers.
It does:: Audits the current branch's changes against eight standing rules, hunting for "code judo": restructurings that keep behavior but make whole branches, helpers or layers disappear.

workflow none none code-review code-quality refactoring

Age: 5 months
Stars: 15.8K; #71
Total Downloads: 23.8K; top 26%
Weekly Downloads: 957; top 29%

deslop

Use it when:: A branch written with an agent is done and you want its machine-written habits cleaned out before review.
It does:: Reads the diff against main and edits out comments, try/catch blocks, any casts and nesting that don't match the surrounding code.

command local none anti-slop code-quality refactoring

Age: 8 months
Stars: 15.8K; #71
Total Downloads: 10.1K; top 48%
Weekly Downloads: 400; top 47%

blast-radius

Part of pstack

Use it when:: A small-looking change makes you uneasy, or you ask what it could break somewhere else.
It does:: Finds the one fact the change's safety rests on, looks where a symbol search can't, and proves that fact with a script or test it runs.

workflow local subagents code-review verification testing

Age: 4 months
Stars: 15.8K; #71
Total Downloads: 7K; bottom 34%
Weekly Downloads: 567; top 32%

how

Part of pstack

Use it when:: You ask how something works, want a walkthrough before changing code, or ask where code should live and which package owns it.
It does:: Sends one read-only subagent for a narrow question; for a subsystem, two to four explorers on separate angles, then one explainer.

workflow none subagents architecture learning

Age: 5 months
Stars: 15.8K; #71
Total Downloads: 6.7K; bottom 33%
Weekly Downloads: 342; top 49%

why

Part of pstack

Use it when:: You ask why code works this way, why a choice was made or what caused a regression, or want its history before changing it.
It does:: Collects the code's git blame, history and PR threads into one brief, then sends one subagent per evidence source it finds, all at once.

workflow none subagents cli architecture research git git github

Age: 5 months
Stars: 15.8K; #71
Total Downloads: 6.4K; bottom 31%
Weekly Downloads: 335; top 49%

technical-writing

Part of pstack

Use it when:: You're writing or reviewing docs, RFCs, READMEs, PR descriptions or commit messages.
It does:: Picks one mode per document (tutorial, how-to, reference or explanation), then applies sentence rules from Google's developer style guide, Simplified Technical English and Global English.

guidelines none none docs-writing writing anti-slop

Age: 3 months
Stars: 15.8K; #71
Total Downloads: 6K; bottom 26%
Weekly Downloads: 387; top 47%

principle-fix-root-causes

Part of pstack

Use it when:: Debugging, and you want the fault fixed at its source rather than guarded where the error showed.
It does:: Reproduces first, asks why at each step until it reaches a wrong decision in the code, and instruments instead of guessing.

guidelines none none debugging

Age: 5 months
Stars: 15.8K; #71
Total Downloads: 5.7K; bottom 21%
Weekly Downloads: 320; bottom 50%

arena

Part of pstack

Use it when:: One attempt at a non-trivial artifact, such as a design or a piece of code, could lock in the wrong shape.
It does:: Writes a rubric, runs candidates in parallel, has a judge from another model family score them, picks a base and grafts the best parts of the others into it.

workflow local subagents multi-agent agent-workflows

Age: 5 months
Stars: 15.8K; #71
Total Downloads: 5.6K; bottom 19%
Weekly Downloads: 326; top 50%

principle-encode-lessons-in-structure

Part of pstack

Use it when:: You catch yourself, or the agent, writing the same instruction a second time, or making the same correction again.
It does:: Encodes the rule as a lint, metadata flag, runtime check or script; when the lesson needs judgment, it makes the instruction more prominent and adds an example of the failure.

guidelines none none code-quality guardrails

Age: 5 months
Stars: 15.8K; #71
Total Downloads: 5.6K; bottom 19%
Weekly Downloads: 316; bottom 50%

principle-prove-it-works

Part of pstack

Use it when:: A task is finished and the agent is about to say it's done.
It does:: Verifies against the real artifact: runs the feature, reads the actual value, inspects the diff, checks process liveness directly.

guidelines none none verification testing

Age: 5 months
Stars: 15.8K; #71
Total Downloads: 5.6K; bottom 19%
Weekly Downloads: 324; top 50%

architect

Only works as part of pstack

Use it when:: You're about to start non-trivial work and jumping straight to code would lock in the wrong shape.
It does:: Grounds itself with how, and with why when ownership or layering changes, has arena runners sketch two or more structurally different designs with stub bodies, then writes the code.

workflow local subagents architecture implementation multi-agent

Age: 5 months
Stars: 15.8K; #71
Total Downloads: 5.5K; bottom 16%
Weekly Downloads: 314; bottom 50%

principle-guard-the-context-window

Part of pstack

Use it when:: Context is filling with large outputs, long files, repeated reads or fan-out planning.
It does:: Routes bulk work to subagents and keeps summaries, not raw payloads, in the main thread.

guidelines none subagents token-saving multi-agent

Age: 5 months
Stars: 15.8K; #71
Total Downloads: 5.5K; bottom 16%
Weekly Downloads: 316; bottom 50%

interrogate

Only works as part of pstack

Use it when:: You want a change or design stress-tested by more than one model before it ships.
It does:: Restates the intent, gives each adversary its own worktree with your uncommitted changes, and keeps a finding only if rerunning its command shows a failure that breaks the intent.

workflow local subagents code-review multi-agent verification

Age: 5 months
Stars: 15.8K; #71
Total Downloads: 5.3K; bottom 9%
Weekly Downloads: 293; bottom 48%

principle-type-system-discipline

Part of pstack

Use it when:: Designing types, reviewing a signature, or writing code in any statically typed language.
It does:: Brands primitives like UserId and OrderId so they can't be swapped, parses external data into typed models at boundaries, and derives types from the schema that owns the shape.

guidelines none none code-quality domain-modeling

Age: 5 months
Stars: 15.8K; #71
Total Downloads: 5.3K; bottom 9%
Weekly Downloads: 334; top 49%

principle-boundary-discipline

Part of pstack

Use it when:: Wiring validation, error handling or framework adapters.
It does:: Validates and handles errors defensively at CLI arguments, config files, external APIs and network protocols, then trusts typed data inside with no re-checks.

guidelines none none code-quality architecture

Age: 5 months
Stars: 15.8K; #71
Total Downloads: 5.1K; bottom 3%
Weekly Downloads: 320; bottom 50%

plugins Skill collection