Skills

Custom Claude Code skills for development workflows. They cover code review, TypeScript fixes, QA automation, PR reviews, design direction, and coding discipline (code reuse + backend performance).

Install

# Install all skills
npx skills add bhagyamudgal/skills

# Install a specific skill
npx skills add bhagyamudgal/skills@done

# List available skills
npx skills add bhagyamudgal/skills -l

Prerequisite

done, audit-ticket and git-commit name a separate unslop skill for the voice of the text they produce. It is not bundled here. Without it those three still apply the punctuation rule each states inline; install unslop to get the rest.

Artifact publishing lives upstream at Folslate and is not bundled here. Install it with npx skills add https://folslate.com/skills/create-artifact.md, plus folslate-api and folslate-cli from the same path when raw HTTP or CLI management is needed.

Update

# Update all installed bhagyamudgal/skills to the latest version
npx skills update

# Or use the upgrade alias
npx skills upgrade

# Update a specific skill
npx skills update bhagyamudgal/skills@review-pr

# Restrict to project- or user-level scope
npx skills update --project    # only project-level skills
npx skills update -g           # only user-level (~/.claude/skills/)
npx skills update -y           # skip scope prompt; auto-detect from cwd

update only refreshes skills you already have. It will not pick up a newly added one. Run add for those.

If anything looks wedged, remove and re-add:

npx skills remove bhagyamudgal/skills -s '*' -y
npx skills add bhagyamudgal/skills

Automatic updates (opt-in)

One paste installs a daily auto-sync: all skills, global scope, every detected agent. It runs add --all (so new skills arrive too, which plain update never picks up) followed by update, at most once per 24 hours. It wires two triggers: a Claude Code SessionStart hook for session starts and an OS scheduler (launchd on macOS, cron on Linux) for background runs.

curl -fsSL https://raw.githubusercontent.com/bhagyamudgal/skills/main/tools/install-skills-sync.sh | bash

Force a sync any time with ~/.local/bin/skills-sync.sh --force. The log lives at ${XDG_CACHE_HOME:-$HOME/.cache}/skills-sync/sync.log. Opt out with:

curl -fsSL https://raw.githubusercontent.com/bhagyamudgal/skills/main/tools/install-skills-sync.sh | bash -s -- --uninstall

Skills (slash commands)

Every row below is one installable skill (skills/<name>/SKILL.md, which is what npx skills add discovers). Non-skill material lives outside skills/ and is listed under Bundled tooling below, so a folder count and the installer always agree. Only flat, public skills are supported: no root or nested SKILL.md, no metadata.internal hidden skills. The verifier rejects those forms.

Skill	Description
done	MANDATORY readiness verification. Map every request item to implementation, verify six acceptance lanes and five evidence facets, and issue final readiness only after required publication evidence exists
simplify	Remove unnecessary complexity from a completed diff while preserving its behavior, guards, and verified scope
parallel-review	Build a reviewer roster, dispatch it in parallel over a local diff, and merge to one ranked list. The merge is not done while any reviewer is outstanding
review-pr	Deep anti-slop review of a GitHub PR with critic-pass filtering, persistent multi-round state, and automatic GitHub submission; self-authored PRs receive the same assessment as a comment
fix-pr-review	Triage and fix CodeRabbit / review-pr findings, then reply + resolve PR conversations
fix-ci	Fix a red check on any PR. Read the failing log first, rerun a flake at most once per run counting earlier attempts, stop when the base is red too, fix forward without force-push or a weakened check, run done, push behind a preflight-mutations card, and stop after two rounds
audit-ticket	Audit a stale GitHub issue against current code, giving per-requirement verdicts with file:line evidence that is re-checked before printing, then update, sunset, or split it
fix-ts-errors	Fix TypeScript errors and loop the workspace type-check until it exits 0. A file whose squiggles cleared is not green
grill-me	Interview-style stress-test of a plan, one decision at a time, then verification of every factual claim against the codebase. No "grill complete" until every decision has an answer and every claim is checked
project-discovery	Discovery interview before writing code on a new project. Interrogate requirements and stack, then emit CLAUDE.md, PATTERNS.md and the lib/ scaffolding
discover-product-domain	Generate through independent naming lenses, remix with the user, and verify their exact standard-price .com domains
design-director	Senior creative-director direction. Eight modes cover brief simplification, staged logo work, layout, typography, color, critique, brand identity, and production asset finalization
browser-qa	Drive a real browser through a UI flow, or exercise an API endpoint directly. Screenshot every step of the flow, check network and console, and account for each step with PASS or FAIL. Tabulate every endpoint case as PASS, FAIL or BLOCKED in a QA Test Results report. For a change headed for a PR, run the same flow on the base code and write a before-and-after fragment for the PR body
reuse-first	Search-first discipline before writing any new utility, type, schema, component, hook, or constant, with a 3-layer search you must print, the reuse ladder, and fork smells
backend-perf	Performance checklist for backend endpoints and DB queries. Walk every check and name a verdict on each. A check you did not name is a check you did not run
systematic-debugging	Four-phase root-cause loop for mid-debugging discipline, where no fix lands without an understood cause, every phase ends on a checkable bar, and bandaid budget is zero
verify-claims	Gate inference-backed, decision-driving claims with a counter-hypothesis and paired evidence from their basis and user-facing acceptance boundary
preflight-mutations	Resolve exact targets, authority, dependencies, reversibility, confirmation, and read-back before changing shared state
bulk-edit	Run one mechanical edit across a whole population of records. Characterize every variant before transforming, refuse the shapes whose meaning is unproven, snapshot and dry run against live, then verify targets, holds and collateral separately
triage-board	Triage a project board in one pass: find candidates, label each agent-ready or need-human, set priority and issue type, and calibrate estimates in human active time against approved anchors. Manual-only via disable-model-invocation, so its description stays out of context until you invoke it
converge-reviews	Reuse review evidence, track affected coverage, and cap stable review scopes at three rounds
manage-report-lifecycle	Consolidate or supersede hosted analytical reports while preserving every source item, one canonical URL, and verifiable predecessor state. Manual-only via disable-model-invocation, so its description stays out of context until you invoke it
executing-tickets-with-subagents	Orchestrate bundled tickets or explicit away/keep-going work through a durable ledger, with one mutation owner per task and a bounded unattended worker pool
sync-agent-setups	Manually preview and sync Claude Code's user-authored behavioral setup outward to explicitly selected agents
resolving-merge-conflicts	Resolve an in-progress git conflict without a silent drop. Every hunk from both sides is placed as kept, superseded, or dropped before you commit
refresh-drizzle-migration	Refresh a branch's stale Drizzle migration after a base merge. Drop it, regenerate it with the project's own drizzle-kit, check its order and that its SQL statements match the old file, then push behind a preflight-mutations card. Never runs a command that touches a database
git-commit	Conventional commits from diff analysis. Every file is classified into exactly one commit. Append-only, with a message-only mode
file-pr	Publish a PR from a current ready-to-publish card, with browser-qa's before-and-after evidence and QA results in the body for UI and API changes, verify the remote branch and PR, then return their evidence to done
file-issue	File one issue an assignee can act on unaided, using two-vocabulary duplicate search, a title that survives a wrong diagnosis, and a body carrying observation, reproduction, expectation and a checkable done condition
openclaw-backup	Verified restore point for an OpenClaw install, with an official archive, VACUUM INTO SQLite snapshots, a raw archive covering the session transcripts the official tool drops, a checksum manifest, and a per-install RESTORE.md. Manual-only via disable-model-invocation, so its description stays out of context until you invoke it
standup	Daily standup update from real activity: GitHub PRs and commits, unpushed local work across worktrees, and Claude/Codex/OpenCode session history

Several skills use progressive disclosure. SKILL.md holds the spine, and branch-specific material sits in references/ (or modes/ for design-director), loaded only when that branch fires. Load instructions use ${CLAUDE_SKILL_DIR}/ so they resolve against the skill directory rather than the user's repo.

Bundled tooling (not slash commands)

Folder	Purpose
coderabbit-config/	.coderabbit.yaml template + persistent-learnings sidecar. Copy into a repo so CodeRabbit absorbs style + convention findings before /review-pr runs. See coderabbit-config/README.md for bootstrap instructions.
tools/verify_skills.py	Structural verifier across all skills. It covers frontmatter, code fences, pointer form, severity-ladder consistency, dangling and orphan references, and cross-skill duplication. Plus produce → validate → consume dataflow checks scoped to review-pr and fix-pr-review. Run python3 tools/verify_skills.py ./skills; exits non-zero on failure.
tools/eval/run_verify_claims.py	Fresh-session behavioral evaluator for verify-claims across code, external mutation, configuration, data, missing evidence, contradiction, and material reversal. Raw streams and final cards are saved under .eval-results/.
tools/eval/run_triggers.py	Routing eval. Runs each utterance in a fresh session and records which skill fires first, so a description change can be checked against every case in tools/eval/triggers.json. Skills that a standing instruction fires in every session, unslop among them, are passed over rather than recorded as the answer.
tools/eval/run_register.py	Register eval for issue #37. Scores prose emitted under a staged skill with slop_score.py. Use --baseline and --variant for isolated trees. --runner codex requires Codex CLI 0.152.1 and a pinned --model, pins high reasoning, ignores mutable user config in an isolated Codex home, and verifies the effective skill catalog before each sample. Summaries record pair-balanced arm order, the normalized Codex context, and protocol_sha256 over the parsers, cases, scorer, rules, and fixture; never combine results from different protocols.
tools/eval/slop_score.py	Scores prose for the AI tells unslop names, plus sentence-length uniformity and nominalisation density. Rules are vendored in slop_rules.json; --check-drift compares them against the live user-local unslop skill.
tools/eval/harness.py	Sandbox construction, transcript parsing and process cleanup shared by the three evaluators. Covered by python3 -m unittest discover -s tools/eval, which runs fully offline.

Usage

/done                # Run after every task
/simplify            # Remove unnecessary complexity without changing behavior
/parallel-review     # Review locally-changed code
/review-pr <pr-url>  # Review a GitHub PR (or several at once in batch mode)
/fix-pr-review       # Triage and apply CodeRabbit / review-pr findings
/fix-ci <pr-url>     # Fix a red CI check on a PR, two rounds at most
/audit-ticket <n>    # Audit a stale issue against current code, then update or sunset it
/fix-ts-errors       # Fix TypeScript errors, loop until the workspace check is green
/browser-qa          # Drive a UI flow in a real browser, or run endpoint QA with tabulated results
/grill-me            # Interview-style plan/design refinement, then fact-check against the codebase
/project-discovery   # Plan a new project
/discover-product-domain  # Generate, remix, and verify product .com names
/design-director     # Design + branding direction
/reuse-first         # Search-first check before writing new utilities/types/components
/backend-perf        # Perf checklist for backend services and DB queries
/systematic-debugging  # Root-cause loop once you're inside a debugging session
/verify-claims         # Verify a consequential inference before relying on it
/preflight-mutations   # Prepare or block a shared-state mutation before execution
/bulk-edit             # Mass-edit one field across many records, with holds, rollback and a collateral check
/triage-board          # Label, prioritize and estimate a board in one pass
/converge-reviews      # Reconcile a review round and choose its bounded next action
/manage-report-lifecycle  # Consolidate hosted reports into one verified canonical artifact
/executing-tickets-with-subagents  # Run a bundled ticket end-to-end via subagents
/sync-agent-setups     # Preview and sync Claude setup to selected agents
/resolving-merge-conflicts         # Resolve merge/rebase conflicts safely
/refresh-drizzle-migration         # Regenerate a stale Drizzle migration after a base merge
/git-commit          # Conventional commit (or message-only)
/file-pr             # Publish a current ready-to-publish card, then return evidence to done
/file-issue          # File one actionable issue, with dedupe search and the same bars
/openclaw-backup     # Verified restore point for an OpenClaw install
/standup              # Daily standup update from real activity

Reference

reference/CLAUDE.md. It is the global user-level ~/.claude/CLAUDE.md these skills plug into. reuse-first and backend-perf were extracted from it into on-demand skills to keep the always-loaded file lean; the rest shows how the skills fit into a full rule setup.

How review-pr + coderabbit-config work together

CodeRabbit is the sieve; /review-pr is the critic-pass. The sieve catches style, convention, and standard-pattern findings so the critic-pass only ever sees what needs judgement, which is intent grounding, codebase-wide reusability (Q6a), multi-round state, and anti-slop filtering on the merged findings.

Adopting coderabbit-config per-repo is what makes /review-pr runs tight. See coderabbit-config/README.md for the per-repo bootstrap.