All authors

Claude Skills by rusharlabs
github.com/rusharlabs34 skills0 installs30 views
- Doc RollupKeeps the project's history/evolution docs (changelog, narrative timeline, state snapshot, lessons, session wrapup) up to date after a significant session — with built-in degradation (a stamp instead of endless narrative) from day 1.Votes: 0GitHub stars: 2
- Pre ClearBefore a /clear, consolidates the LONG TERM (conditional doc-rollup — how we got here) and the SHORT TERM (handoff — what comes next), then renders the BOOT BUNDLE.Votes: 0GitHub stars: 2
- Gotcha MemoryOperational learning loop — records command failures, detects recurrence and injects the lesson as a preamble before the next execution of the same task. Use to query/seed/debug the project's gotcha memory.Votes: 0GitHub stars: 2
- Architecture Decision RecordsCaptures architectural decisions made during the session as structured ADRs (context, alternatives considered, consequences) in docs/adr/. Use when the user decides between significant alternatives (framework, database, pattern) or asks "why did we choose X?".Votes: 0GitHub stars: 2
- Claude Dev SetupInstalls the base Claude Code hooks in a new project - idempotent and reversible settings wiring (never overwrites someone else's config without --force) + chain-preserving git hook (never replaces an existing pre-commit hook).Votes: 0GitHub stars: 2
- HookifyCreates REAL hooks for Claude Code — executable scripts (stdin JSON, exit 0/1/2), registered via a plugin's hooks.json or pasted into settings.json. Use when the user wants to create a hook, a safety rule, a custom validation, a lifecycle hook.Votes: 0GitHub stars: 2
- Plugin DevPackages skills/hooks/commands into an installable Claude Code plugin — the real anatomy (.claude-plugin/plugin.json + hooks/hooks.json + ${CLAUDE_PLUGIN_ROOT}), no framework/build step. Use when the user wants to create a plugin, package an extension, or distribute a set of skills/hooks.Votes: 0GitHub stars: 2
- Search FirstSearches for an existing library/tool/pattern BEFORE writing new code - covers package registries (npm/PyPI), MCP and GitHub, in addition to the local grep. Use before creating a utility, helper, or new integration.Votes: 0GitHub stars: 2
- Skill ScoutSearches for skills locally, in the marketplace, on GitHub and on the web BEFORE creating a new skill — avoids duplicating work that already exists. Use when the user says "create a skill", "is there a skill for X?", or you are about to suggest creating a new skill.Votes: 0GitHub stars: 2
- Skill WriterGuides the creation of Agent Skills for Claude Code — structure, frontmatter, effective descriptions, and validation against the SKILL-CONTRACT. Use when the user wants to create, write or structure a new skill.Votes: 0GitHub stars: 2
- TeachingTurns any technical output (creation, structure, architectural decision) into a learning opportunity - a tree of where the element lives, an x-ray of what-it-is/where-it-sits/what-it-is-for, a connection map, a business analogy, and explained decisions. Always use in technical output for a non-programmer reader.Votes: 0GitHub stars: 2
- Pp ConsolidateParallel Process Consolidate - consolidates the outputs of parallel sessions/agents into a single, deduplicated, verifiable verdictVotes: 0GitHub stars: 2
- Pp DiscoveryParallel Process Discovery - token-safe inventory of repositories, folders and large artifacts before deep analysisVotes: 0GitHub stars: 2
- Pp XrayParallel Process X-Ray - line-by-line reading of an external module/repo with evidence, risks and calls for the next investigationVotes: 0GitHub stars: 2
- Dashboard BuilderBuilds monitoring dashboards (Grafana, SigNoz and similar) that answer real operator questions, not "show every metric that exists". Use when turning a list of metrics into a genuinely operable dashboard.Votes: 0GitHub stars: 2
- Health CheckProbes a config-driven list of services (HTTP or local command) and writes a JSON cache that another tool (e.g. the statusline) can read without touching the network — never on the hot path itself, it only generates the cache.Votes: 0GitHub stars: 2
- Adversarial RefuterBefore accepting "done/ready", dispatches read-only refuters that try to KNOCK DOWN the claim against the disk/live sourceVotes: 0GitHub stars: 2
- Claude Md From ProfileGenerates the project's CLAUDE.md block FROM operator-profile.yaml - floor first (what the AI does not decide), then where things live, the rules of done and the flow. Idempotent, signed; refuses to overwrite a hand-edited blockVotes: 0GitHub stars: 2
- Delegate With HandbackDelegates a long/independent task to a 2nd agent with explicit context and a verification gate on the returnVotes: 0GitHub stars: 2
- Doc Consolidator DedupMerges N overlapping docs/plans into a single deduplicated work-list, archives the superseded ones with a stub-redirect — grep/ls before creating/classifying (LC-3)Votes: 0GitHub stars: 2
- Dual Report BuilderGenerates TWO versions of the same analysis/report — INTERNAL (raw, failures exposed, dark) and EXTERNAL (premium, positive, no failures exposed, light) — as self-contained HTML with pure-CSS charts, print-friendly. The external version goes through a sober-register gate (banned phrases from the profile). Use when producing a report/dashboard with a dual audience (internal team + client/stakeholder).Votes: 0GitHub stars: 2
- Gate Sheet CollectorDrains everything that depends on the human into ONE form (exact command + what-it-unblocks), without ever blocking the loopVotes: 0GitHub stars: 2
- Gated Improvement ProposalEvery self-edit of the harness (rule/CLAUDE.md/prompt/hook) becomes a PROPOSAL that passes a gate before being applied — never a sensitive auto-mergeVotes: 0GitHub stars: 2
- Live Source ProverBefore citing ANY number/status/metric, re-derives it at the live source and labels it "live @ HH:MM + source" — never repeats stale dataVotes: 0GitHub stars: 2
- Parallel DispatchFires independent tasks in waves with a concurrency ceiling + sequential fallback on rate limitVotes: 0GitHub stars: 2
- Pre Clear Boot BlockBefore a /clear, emits the DONE / MISSING / read-in-this-order block so the fresh session resumes with zero lossVotes: 0GitHub stars: 2
- Rls AuditReally audits the RLS of a Supabase project — pg_policies per permissive anon policy + get_advisors, not just the relrowsecurity flagVotes: 0GitHub stars: 2
- Supabase Edge ScaffoldScaffolds a Supabase Edge Function with correct CORS + service-role + error handling, instead of copying boilerplate by handVotes: 0GitHub stars: 2
- Rls AuditAudits a Supabase project's RLS for real — pg_policies per permissive anon policy + get_advisors, not only the relrowsecurity flagVotes: 0GitHub stars: 2
- Supabase Edge ScaffoldScaffold of a Supabase Edge Function with correct CORS + service-role + error handling, instead of copying boilerplate by handVotes: 0GitHub stars: 2
- Lane CoordinatorCoordinates N concurrent sessions (lanes) over the same repo through a whiteboard with a state machine (lane_board.py) — from CLAIMED to MERGED, with cross-model maker≠checker enforced in code, not by textual discipline.Votes: 0GitHub stars: 2
- Agent Framework ScaffoldRuns the 6-step wizard that generates the skeleton of a new project (operator-profile.yaml + the chosen templates) — check environment → configure → validate → generate.Votes: 0GitHub stars: 2
- House SessionRuns a House Session on this host — a panel of pinned deciders from at least two model families, each seat a read-only command in its own lane, blind first round, counted, stopped and sealed by the HPP core.Votes: 0GitHub stars: 2
- Loop DriverTurns the agent into an autonomous lead engineer — reads charter+work-list, executes until exhausted, self-prompts, stops at the stop-conditionsVotes: 0GitHub stars: 2