Iteratively trims a Claude Code skill (SKILL.md + its bundled resources) down to exactly what it needs — repeated review/fix rounds against a concrete duplication-and-bloat checklist, with a fresh independent reviewer each round, until two consecutive rounds find nothing left to change (or a hard cap is hit and reported as non-convergence, never silently accepted). Bundles the skill-tighten skill and its read-only skill-tightener-reviewer agent.
Plugin skill-tighten from marketplace umbraco-mcp-ops in hifi-phil/umbraco-mcp-ops. Commands read from the repo's manifest — 19 days ago.
skill-tightenIteratively trims a Claude Code skill down to exactly what it needs: repeatedly reviews the skill's entire directory against a concrete checklist, then applies the fixes, repeating until two consecutive rounds find nothing left to change — or a hard iteration cap is hit and reported as non-convergen
lossless-reviewerFresh, read-only antagonist for skill-tighten's post-convergence gate. Given a skill's pre-tightening snapshot and its current (tightened) state, checks every line the tightening rounds removed for whether the same substantive information survives somewhere in the final skill — flagging anything tha
skill-creator-validatorFresh reviewer for skill-tighten's post-convergence gate. Audits a finished skill against skill-creator's own "Skill Writing Guide" (anatomy, progressive disclosure, description quality, writing style) to catch structural/quality issues the tightening checklist doesn't check for, since that checklis
skill-tightener-reviewerFresh, read-only reviewer for skill-tighten's review-fix loop. Given a skill directory, judges it against the tightening-checklist and returns concrete findings (each with an exact location and an exact fix) or a clean pass. Read-only — cannot edit; the loop applies whatever it finds.
Label-driven loops that take an issue from rough idea to merged PR: `ai-discussing` talks it into shape before anyone builds it, `ai-ready` builds it to a PR, `auto-reworking` acts on your review feedback. Bundles issue-discuss-loop, issue-build-loop, rework-loop, and mcp-review. Fired by the agent-orchestration Worker as a cloud routine. Repo-agnostic.
5 skills · 2 agents
Branching, merge, release, and dev-sync workflow skills for any repo — detects gitflow (dev + main) vs main-only and follows the matching conventions. Bundles release-and-branching, sync-dev, and auto-release-loop (issue+label-triggered release with two automated gates — green CI + an Opus pre-publish review against a growing checklist — no approval pause).
4 skills · 1 agent
Shared skill for writing the structured outcome artifact a loop appends to a comment it already posts, so the agent-orchestration Worker can read what happened and apply the labels itself. One marker format, one growing catalog of outcome shapes; loop plugins point here instead of each inventing their own. Loops report through post-outcome.sh, which signals the agent-orchestration Worker and prints the exact comment body to post. Bundles hooks that signal it too: heartbeats naming the step while a loop works, blocked on a permission prompt, and ended when a session stops without reporting (AGENT_OUTCOMES_ENDPOINT + AGENT_OUTCOMES_TOKEN; without the endpoint they send nothing). Also bundles the work-log skill: how a loop records its decisions and what it verified in the Worker's decision log and build log, with log-entry.sh to add and read entries.
2 skills · hooks
Single front door for a repo's automation loops so one routine per repo can handle every loop event instead of one routine per event. Deterministically routes each triggering GitHub event to the matching loop (via route-event.sh); a non-matching event is a quiet no-op. Router only; each loop owns its guardrails. Also bundles new-loop-routine — the standard config + canonical prompts + UI wiring for standing up one routine per repo. Requires github-ops + the loop skills.
3 skills
The self-learning system shared by every automated loop in this repo. SubagentStop/SessionEnd hooks capture proto-learnings — raw observations worth improving — as rows on a shared Slack canvas (not GitHub issues, which silently lose captures across an org boundary). The hooks fire for any loop, matching on shared conventions (github-ops, /goal) rather than naming individual loops, so a new loop needs no integration here. Bundles two separate skills: proto-learning-capture (reference for the capture half and its row schema) and triage-learnings (Loop B), which weekly dedupes/clusters the canvas inbox and routes each cluster to a tracked issue on the owning MCP repo, a gated PR to the shared umbraco-mcp-skills, or a loop-improvement issue on this repo. Install alongside any loop plugin to get capture for free.
2 skills · hooks
Two halves, deliberately separate. The branch-housekeeping SKILL reports the state of the remote branches across every configured Umbraco MCP repo in one run and posts a Slack digest — classifying each non-protected branch by its GitHub PR state (authoritative where git ancestry isn't, since these repos squash-merge) into safe-to-remove, merged-but-reused-since, closed unmerged, no PR, and open. It is informational and read-only — never deletes — so it needs no write scope and runs as a scheduled CLOUD routine as well as locally. The /clean-branches COMMAND is the separate, deliberate cleanup: a purely mechanical script with no model in the loop, which re-verifies every branch against the API — protected lists, live default branch, live protection, merged_at, and that the tip still matches what the PR merged — before deleting it. Run it only when you want branches gone. The command is LOCAL-ONLY — the GitHub MCP server exposes no branch-delete tool, so cleanup cannot be scheduled at all. Also flags any repo whose "Automatically delete head branches" setting is off, which is the durable fix cleanup only works around.
1 skill · 1 command
Matt Pocock's agent skills for real engineering (grilling, spec/ticket flows, TDD, code review, domain modelling and more).
39 skills
Core skills library for Claude Code: TDD, debugging, collaboration patterns, and proven techniques
15 skills · hooks
Official Next.js skills: adopt and optimize Cache Components, adopt Partial Prefetching, and verify runtime behavior against a running dev server.
6 skills
tasteskill 2.5.1: professional company websites at the figma.com and x.ai level, with one section per screen, real photographs, real product UI, effects kept for tech and gaming pages, and a real-browser checker and finish gate.
1 skill
5 skills + 35 modes + Material Passport pipeline. Includes v3.6.7 cross-model audit gate and v3.6.8 generator-evaluator contract.
5 skills · 16 commands · 4 agents · hooks
Claude Skills for Obsidian
6 skills
This page updates itself from your repo. Add the badge to your README, or get it Featured to pin it above the organic results with a followed link to your site.