You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Added a Simplicity (over-engineering) review dimension and a Simplify-focus delete-list mode (delete/stdlib/native/yagni/shrink, net-lines metric) to the reviewing-code skill.
Plugins
Plugin
Description
browser
Browser-based testing, validation, screenshots, recordings, and quick automation
dev-flow
Code review, fixes, commits, and linting for development workflows
discovery
Research, docs lookup, instruction review, reasoning, and agent configuration audits
Added turnkey Playwright screenshot helpers for single URLs and URL-template sequences, including JSON manifests with titles, screenshot paths, console errors, network failures, and bad HTTP responses.
Added regression tests for the Playwright runner and screenshot helpers.
Fixed
Hardened the Playwright support runtime so scripts can import Node modules or Playwright explicitly while still receiving injected browser globals.
Kept runner status logs off stdout for reliable JSON output, preserved caller working directories, and stopped writing execution temp files into the skill directory.
Improved browser automation guidance for Pi and headless harnesses to prefer screenshot and manifest evidence when no visible browser is exposed.
Plugins
Plugin
Description
browser
Browser-based testing, validation, screenshots, recordings, and quick automation
dev-flow
Code review, fixes, commits, and linting for development workflows
discovery
Research, docs lookup, instruction review, reasoning, and agent configuration audits
Collapsed the spec-driven workflow from seven public spec skills into one lightweight spec-flow skill that plans one slice, executes one task, checkpoints or closes, and repeats.
Replaced the multi-skill specctl helper with a smaller state-only CLI for .spec/ artifacts, readiness, sessions, checkpoints, validation, dependencies, and completion evidence.
Updated spec-flow documentation, method guidance, README catalog counts, and generated Claude, Codex, Gemini, and Pi artifacts for the 25-skill catalog.
Removed
Removed the separate spec-init, spec-interview, spec-plan, spec-new, spec-status, spec-work, and spec-done skills from the installable spec-flow plugin.
Fixed
Updated spec-related skill-enforcer routing and tests to match the single spec-flow skill while keeping neighboring skills independent.
Plugins
Plugin
Description
browser
Browser-based testing, validation, screenshots, recordings, and quick automation
dev-flow
Code review, fixes, commits, and linting for development workflows
discovery
Research, docs lookup, instruction review, reasoning, and agent configuration audits
Reworked documenting-code around reader-aware documentation updates: human docs, agent-facing docs, code comments, and language-specific documentation guidance now stay grounded in code facts, use smaller doc deltas, and apply tighter Claude tool and workflow limits.
Reworked brainstorming-ideas to prefer interactive question tools over plain-text numbered menus, tightened the grill protocol, reduced the Claude overlay to platform-specific deltas, and added paused/routed output handling.
Reworked deploying-infra into a validate-first deployment workflow with explicit apply confirmation for every environment, validate-only handling for Dockerfiles and GitHub Actions, repo-detected deploy commands, stronger blocked/failure outputs, and more realistic validation checklists.
Reworked fixing-code so the Claude overlay and tool envelope now support read/edit/write/questioning flows, scoped one-defect fixes, blocked/proposed-change handling, and tighter route-away rules for browser-only debugging.
Reworked improving-tests so the Claude overlay can actually apply test changes, emits blocked/proposed-change outputs, follows convention-first language references, and uses leaner Go, Python, TypeScript, and Web test guidance without absolute rules.
Reworked researching-web around a single canonical Research Result output contract, task-driven web-tool selection, stronger source-grounding, explicit source-conflict handling, and stale-source-risk reporting.
Updated README and AGENTS catalog copy to match the 31-skill suite and the revised deployment and research skill scopes.
Removed
Removed the low-use exploring-repos skill and DeepWiki routing. Public repo architecture questions now fall back to normal web research, docs lookup, GitHub CLI, or local exploration instead of a dedicated skill.
Fixed
Fixed the README skill-count badge so it matches the current 31-skill catalog.
Fixed instruction/tool drift across the updated skills by regenerating all target-specific dist/ artifacts from src/.
Plugins
Plugin
Description
browser
Browser-based testing, validation, screenshots, recordings, and quick automation
dev-flow
Code review, fixes, commits, and linting for development workflows
discovery
Research, docs lookup, instruction review, reasoning, and agent configuration audits
Reworked evolving-config into a review-first AI coding-agent configuration audit workflow: fixes now require an explicit request, --fix, or named approval before changing permissions, sandbox policy, hooks, MCP servers, model routing, package installs, private config, or managed settings.
Added platform-specific guidance for Claude Code, Codex, and Pi configuration surfaces, including settings, context files, skills, agents, hooks, MCP, sandbox and approval policy, packages, extensions, prompts, and generated-export rules.
Tightened the evolving-config rubric around evidence, least privilege, generated-file source of truth, routing precision, context efficiency, and validation readiness.
Updated the Claude overlay to use local config evidence first, official docs and changelogs second, and broad web research only for gaps or ecosystem comparisons.
Updated the discovery plugin and catalog descriptions to describe research, docs lookup, repo exploration, instruction review, and agent configuration audits.
Fixed
Removed Gemini from evolving-config best-practice scope; explicit Gemini requests now get local-file review only with the coverage gap stated.
Removed stale audit/apply mode conflicts so review-only prompts no longer drift into fix prompts.
Plugins
Plugin
Description
browser
Browser-based testing, validation, screenshots, recordings, and quick automation
dev-flow
Code review, fixes, commits, and linting for development workflows
Tightened refactoring-code into a focused batch-refactor workflow with
explicit route-away rules, behavior-preservation gates, mapped-site evidence,
public API compatibility guidance, and optional GitNexus/codegraph impact
checks for renames, moves, extracts, and broad restructures.
Reworked fixing-code around a reproducible feedback loop: reproduce first,
diagnose with evidence, patch the smallest root cause, add regression coverage
where a real seam exists, and clean up temporary instrumentation before
reporting success.
Reworked improving-tests around behavior seams, characterization tests,
TDD slices, coverage-as-signal guidance, flaky/slow/dead-test review checks,
and optional GitNexus/codegraph evidence for affected flows and high fan-in
regression surfaces.
Upgraded reviewing-code with quick, standard, deep, team, and external modes;
a dedicated severity/confidence rubric; explicit score caps; Needs review
handling for missing context; consolidated subagent/external-review rules; and
focused GitNexus/codegraph guidance for PR blast radius and missed caller or
test coverage checks.
Rebuilt reviewing-instructions around a canonical 0-10 band-first scoring
rubric with hard gates, caps, confidence, calibration anchors, model alias
resolution, and scoped lint pre-pass support for more stable repeated scores.
Simplified review language references for Go, Python, TypeScript, and Web so
they are read-only compatible, tool-aware without being tool-first, and focused
on concrete defect classes instead of tutorial-style examples.
Fixed
lint-instructions.py now honors explicit file, directory, and plugin scopes
instead of reporting unrelated whole-repo findings during targeted instruction
reviews.
The instruction-scoring rubric and model references now pass their own format
lint rules, avoiding self-reported horizontal-rule, italic, and bold-overuse
findings.
README and AGENTS skill catalogs now describe the updated review, debugging,
testing, refactoring, and instruction-scoring workflows.
Plugins
Plugin
Description
browser
Browser-based testing, validation, screenshots, recordings, and quick automation
dev-flow
Code review, fixes, commits, and linting for development workflows
discovery
Research, reasoning, config evolution, and agent self-improvement
Reorganized the installable plugin catalog from eleven source plugins to
seven coherent plugins: browser, dev-flow, discovery, git-flow, infra-ops, programming, and spec-flow.
Renamed browser-automation → browser, dev-tools → discovery, and spec-dev → spec-flow, and consolidated go-dev, py-dev, ts-dev,
and web-dev into programming.
Added declarative plugin dependencies in source manifests: browser → programming, programming → dev-flow, and spec-flow → dev-flow.
Merged reviewing-cc-config into evolving-config, keeping config review
and config-improvement workflows in one skill with review-only and
apply-fixes modes.
Clarified cc-thingz as “Coding Companion” in the top-level docs and reduced
Claude-first framing in the README and AGENTS catalog.
Kept revdiff-plan-review as an optional Pi bridge hook and assigned it to spec-flow instead of leaving it orphaned.
Removed
Removed the empty test-e2e plugin.
Removed thin or redundant skills: analyzing-usage, learning-patterns, mem-history, parsing-documents, and smart-explore.
Removed dead skill-enforcer routes for deleted or phantom skills: learning-patterns, mem-history, reviewing-cc-config, smart-explore, using-gemini, and coding.
Fixed
Updated README, AGENTS, contributing docs, release-mirror rewrite rules,
generated manifests, and hook/test fixtures to match the new plugin layout,
names, counts, and configuration surface.
Regenerated all marketplace manifests, plugin outputs, and target-specific
artifacts for the 5.0.0 layout.
Plugins
Plugin
Description
browser
Browser-based testing, validation, screenshots, recordings, and quick automation
dev-flow
Code review, fixes, commits, and linting for development workflows
discovery
Research, reasoning, config evolution, and agent self-improvement
Renamed the user-facing browser workflow from testing-e2e to browser-automation, broadened it to rendered UI exploration, validation,
screenshots, recordings, and browser-flow tests, and made playwright-skill
support-only.
Slimmed playwright-skill from a Playwright tutorial into a runtime-only
helper reference, fixed dev-server detection, and aligned temporary scripts
and artifacts under /tmp/playwright-*.
Folded debating-ideas into brainstorming-ideas and retained the
challenge protocol as a conditional reference.
Consolidated Context7 command guidance into looking-up-docs, keeping docs
lookup in one user-facing skill with official-source and web fallback
references.
Tightened runtime-neutral instruction wording across several support skills
and generated targets.
Plugin and marketplace versions bumped to 4.12.0.
Removed
Removed standalone testing-e2e, debating-ideas, and context7-cli skill
entries after their behavior moved into the replacement skills above.
Plugins
Plugin
Description
browser-automation
Browser automation for rendered UI exploration, validation, screenshots, and end-to-end flows
dev-flow
Code review, fixes, commits, and linting for development workflows
dev-tools
Modern CLI tools, git worktrees, docs lookup, research, and brainstorming
go-dev
Idiomatic Go development with stdlib-first patterns
infra-ops
Kubernetes, Terraform, Helm, GitHub Actions, AWS, and GCP operations
py-dev
Python 3.12+ with uv/ruff/pyright toolchain and modern patterns
spec-dev
Spec-driven development with structured requirements and planning
ts-dev
TypeScript with strict typing, React patterns, and modern tooling
web-dev
Web frontend with vanilla HTML, CSS, JavaScript, and HTMX
Tightened writing-go instructions and Go references for leaner scope,
version-gated APIs, dependency-aware concurrency guidance, and lowercase
reference filenames.
Plugins
Plugin
Description
dev-flow
Code review, fixes, commits, and linting for development workflows
dev-tools
Modern CLI tools, git worktrees, docs lookup, research, and brainstorming
go-dev
Idiomatic Go development with stdlib-first patterns
infra-ops
Kubernetes, Terraform, Helm, GitHub Actions, AWS, and GCP operations
py-dev
Python 3.12+ with uv/ruff/pyright toolchain and modern patterns
spec-dev
Spec-driven development with structured requirements and planning
test-e2e
End-to-end testing with Playwright for browser automation and UI testing
ts-dev
TypeScript with strict typing, React patterns, and modern tooling
web-dev
Web frontend with vanilla HTML, CSS, JavaScript, and HTMX