Cxc Dev logo

Cxc Dev

Community
lidge-jun
cxc-dev

Use for coding, PR delivery, scaffolding and QA. Triggers: develop, fix, refactor, test, review, docs, browse, stacked PR, 개발, 수정, 검토, 스택 PR.

Overview

Publisherlidge-jun
Repositorycodexclaw
Skill namecxc-dev
Stars
51
Forks
10
Bundled files
17
LicenseMIT
Links
  • Markdown instructions

    A SKILL.md file the model loads on demand, so it only costs tokens when a request actually matches.

  • Works with any LLM

    AI skills are plain Markdown, not provider-specific code, so this works with GPT, Claude, Gemini, Grok, or a local model.

  • 17 bundled files

    Scripts, templates, and references the model can read while it works. Files are read-only and never executed.

  • Open source

    Published by lidge-jun on GitHub. Read the source before you install it.

Installation

Install the Cxc Dev AI skill in TypingMind to use it with any LLM, or drop it into another agent that reads SKILL.md.

1

Install in TypingMind

TypingMind installs a skill straight from its GitHub folder — it reads SKILL.md, bundles the resource files, and stores the result locally.

  1. Open the app and go to Plugins → Skills.
  2. Choose "Install from GitHub".
  3. Paste the skill folder URL below and confirm.
    https://github.com/lidge-jun/codexclaw/tree/main/plugins/codexclaw/skills/dev
  4. Enable the skill in any chat where you want it available.
Plugins → Skills → Add skill → From GitHub URL, then paste the folder URL and press Continue.
2

Install in another agent

Any agent that reads the Agent Skills format can use this skill — copy the folder into that agent's skills directory.

Claude Code — .claude/skills
git clone --depth 1 https://github.com/lidge-jun/codexclaw.git /tmp/codexclaw
mkdir -p .claude/skills
cp -r /tmp/codexclaw/plugins/codexclaw/skills/dev .claude/skills/cxc-dev
Restart Claude Code after copying so it picks up the new skill.

Use it in TypingMind

Enable Cxc Dev in any TypingMind chat and the model takes it from there. Its name and description sit in the system prompt, and the moment a request matches, the model loads the full instructions itself — you never invoke it by hand, and it costs no tokens until it is actually used.

The model loads Cxc Dev on its own as soon as a request matches it.

Works with any AI model

AI skills are plain Markdown instructions rather than provider-specific code, so Cxc Dev is not tied to the model it was written for. Install it once in TypingMind and use it with GPT-5, Claude, Gemini, Grok, DeepSeek, Mistral, Llama, or a local model you run yourself — all on your own API keys.

  • Loaded only when it is needed

    The system prompt carries just the name and description. The instructions are fetched on the first matching request, so an idle skill costs nothing.

  • Switch models mid-chat

    Because the skill is instructions rather than code, changing model does not break it — the next model reads the same SKILL.md.

Skill instructions

This is the SKILL.md content the model loads. Read it before installing — a skill is instructions your model will follow.

Dev

User/host safety and tool contracts prevail. Diagnosis/review permits investigation only; changes stay in authorized scope. Prose is not runtime enforcement.

§0.0 Work Classifier (C0-C5)

DEV-CLASS-01 (DEFAULT): classify before depth; reclassify growth. Decide discovery ownership before broad reads, even read-only.

ClassSignalsProcess
C0Text only; zero behavior changeDirect edit + smallest proof
C1One file, local behavior, no new abstractionFast path + focused check
C2Conventional product sliceCompact plan, adjacent conventions, focused tests, micro-audit
C3Module boundaries, shared types, broad behaviorPABCD depth by risk/persistence; subagent audit when warranted
C4Auth, payments, deletion, migration, release, permissions, securityFull PABCD, gates, durable risk/evidence
C5Ambiguity after one clarification roundInterview via pabcd; resolve/reclassify C0-C4 before build

Tie-break: higher class wins (DEFAULT); route→service→storage remains C2. DEV-ESCALATE-01 (STRICT): security, deletion/migration, destruction, public contracts, release, permissions, new dependencies/frameworks promote the affected part to C4. Split it out; promotion alone requires no question.

§0.1 Patch Fast-Path (C0/C1)

C0/C1 skip convention/owner search and references unless explicitly routed; keep §3/4/5/static checks and visible conventions. Five lines is illustrative. Security/data-loss/new abstractions exclude this path; text-only edits do not escalate by location. C0 needs no numbered unit; C1 records change/reason/proof in existing units only. Never create units just for C0/C1; unit residence owns UNIT-RESIDENCE-01.

§0.2 Rule Classes

Safety/correctness/permissions/truthful proof are mandatory. Size/naming/layout/style are DEFAULT/STYLE_SAMPLE despite old MUST/HIGH labels unless a cited project/user contract binds them. Requested workflows keep phase/evidence; other unclassified rules are DEFAULT.

  • STRICT: violation blocks completion.
  • DEFAULT: follow unless a stated, documented reason warrants deviation.
  • HEURISTIC: judgment guide; deviation needs no rationale.
  • DEV-STYLE-SAMPLE-01 (STYLE_SAMPLE): illustration, never universal.
  • ESCALATE: stop and ask the user.

§0.4 Workflow Modes

Mode: chat, PABCD, goal, scoped subagent, read-only (no writes), docs-only (consistency checks). Read pabcd/cxc-loop for mechanics; classify each work-phase. Docs-first: cxc-loop (LOOP-DOCS-FIRST-01). Production: deployed to real users beyond the author; excludes prototypes/spikes/internal demos.

Companion Skills

DEV-ROUTE-01 (STRICT): before writing, read every matching surface router SKILL.md; C0/C1 exceptions apply.

Reading contract

Read active SKILL.md fully; references only when applicable. Resolve a skill's relative links from that skill's directory, not the working directory. Truncation is incomplete even at exit 0: reread separately or in numbered contiguous, non-overlapping chunks through EOF; verify no gaps before acting.

Change surfacePrimary routerAlso load
Backend / API / serverdev-backenddev-security for auth/input
Frontend / UI / webdev-frontenddev-uiux-design for vague/open visual direction, UX-state meaning, IA, brand, concept gen
App database / OLTP / transactional schemadev-backenddev-security for access; dev-testing for migrations
Analytics / ETL / data quality / analytical backfillsdev-datadev-backend for API integration
Tests / QAdev-testingdev-frontend for browser QA
Security / auth / secretsdev-securitysurface-specific router
Architecture / modules / depsdev-architecturedev-scaffolding for new structure
Debugging / crashes / perf / comprehending unknown systemsdev-debuggingsurface-specific router; references/logic-analysis.md when no defect
DevOps / deploy / infradev-devopsdev-security for credentials
Native desktop: Tauri, AppKit/SwiftUI, WidgetKit, menu-bar/tray apps, embedded sidecars, signing/notarization, macOS approval promptsdev-devops → references/native-desktop-acceptance.md and references/macos-system-approvals.mddev-testing for suites and CI matrix; cxc-qa for gui/cli verdicts; dev-frontend for web-view UI; dev-security for entitlements and credentials; dev-debugging runtimes/swift.md
Scaffolding / docs / setupdev-scaffoldingdev-architecture for boundaries
Code reviewdev-code-reviewerdev-security + dev-testing
Diagrams / charts / visual documents / reports / PDF compositiondev-visualizerAvailable document-format owner for PDF/DOCX/Slides mechanics; dev-frontend and dev-uiux-design retain implementation/design ownership

Subagent Skill Injection (DEV-SKILL-INJECT-01)

Attach dev+surface skills explicitly; search tasks also attach search and its policy. Use resolvable links or supported mentions/v1 items; hooks never infer omissions. Mappings: ownership; triggers: agents/openai.yaml.

Discovery delegation

Before broad reads, bound the question; delegate to configured explorer if independent and main has separate work. State scopes; host/no-delegation limits win. Record unavailability; read narrowly. Local lookup must determine next steps or be inseparable; explain substantial reads. Read-only/counts/parallelism do not justify them. Revisit on new subsystems/rereads/truncation; reclassify growth. Use a discovery packet; check needed spans, reassign named gaps before expanding, no routine re-investigation. Discovery replaces neither implementation delegation nor independent review. Isolation: dispatch surfaces; read-only leaves do not justify parallel writes. Authorized delegation needs no new approval. Model claims need runtime proof; cost compares main+child input/cache/output prices/tiers, not tokens.

Family Invariants

  • FAMILY-SLOP-01 (DEFAULT): no filler, performative narration, decorative rationale, placeholders, TODO-only output, fake fallbacks, speculative wrappers/defensive clutter without boundary reasons. Prose/page tells: kwrite and dev-visualizer REPORT-DESIGN-01.
  • FAMILY-READER-01 (DEFAULT): reader documents for human deliverables; link raw audit artifacts.
  • FAMILY-CITE-01 (STRICT): findings/plans/reviews/contradictions cite path:line; plans name paths/verifiers; proof includes command+output/artifact.
  • FAMILY-PROOF-01 (STRICT): completion follows §3; all routers inherit it.

3. Verification Before Completion (STRICT)

  1. Identify the command proving the claim.
  2. Run it fresh, never cached.
  3. Read full output; check exit code and count failures.
  4. Confirm it supports the claim.
  5. Report the claim with evidence.
ClaimRequiresInsufficient
Tests passExpected tests executed; fresh zero failuresOld/absent/skipped/cancelled tests, assumptions
Build succeedsBuild exit 0Lint success
Bug fixedOriginal symptom resolvedEdit alone
Feature completeEvery requirement checkedTests alone
Subagent completedActual changes in VCS diffChild report alone
Regression test worksVerified red→greenOne passing run

DEV-VERIFY-FLOOR-01 (STRICT): floor, not cap:

ClassFloor
C0/C1C0 text consistency; C1 focused check, or observed repro with stated limits when automation does not fit
C2Integration/contract tests + build/typecheck + changed-UI smoke; CRUD negatives: dev-testing references/core/crud-test-matrix.md
C3Affected suites + public-contract/doc consistency
C4Full relevant gates + negatives + durable evidence

Verify child success independently: diff and behavior.

4. Change Documentation

C2+ supplied logs record change/reason/proof. C0/C1 follow §0.1; discovered logs create no duty. Explicit requests/release contracts bind named logs; no unrelated records.

5. Safety Rules

  • Search consumers before export removal; unused internal exports may go within scope; public removal needs compatibility/migration decisions.
  • Confirm import target files/exports; externalize config in files/env and name magic constants.
  • Surface async errors: verified JS/TS backend neverthrow boundaries (dev-backend §3), otherwise try/catch with contextual logs.
  • Destruction (ESCALATE): explicit approval for file deletion/table drops/state resets/cache clearing.
  • DEV-GIT-COMMIT-01 (DEFAULT): locally commit each complete implementation step; do not accumulate a feature uncommitted.
  • DEV-GIT-PUSH-01 (ESCALATE): explicit current-session approval for remote pushes, force pushes, remote branch creation and tags. Local commits are autonomous; completion grants no push permission.
  • DEV-SHELL-TEXT-01 (STRICT): generated prose/replacements use apply_patch/files/quoted-delimiter heredocs/--body-file, never double-quoted shell or sh -c payloads. Use git -C <source>; hooks cannot see workdir. SHELL-SUBST-01/WORKTREE-GUARD-04 are narrow safeguards.
  • DEV-PRIVACY-01 (STRICT): no raw client/personal transcripts, documents/figures, named private conversations or credentials in repo evidence/fixtures/samples. Store outside checkout; devlog records filename/date/summary. Codexclaw plans and evidence follow cxc state path (outside the checkout in external mode). Before first branch push, list identifying company/product/people/distinctive-number terms and search the push range (DEFAULT). Fix hits by rewriting unpushed commits, never follow-up commits that retain exposed history.
  • Dependent delivery follows stacked PRs (DEV-STACK-01, DEV-STACK-02, DEV-STACK-04, DEV-STACK-OPT-IN-01); retain its merge/native-selection grants.

Conditional Routes

Read matching rows; repo tools own local facts.

WhenRead
PR creation/review/merge, dependent deliveryStacked PRs (DEV-STACK-06/07)
Method/repo requirement/strict triggerMethodology overlays
C2 conventional sliceCRUD
C2+ implementation/new abstractionPractice
Debug/comprehendcxc-dev-debugging; no defect: its references/logic-analysis.md
Hosted CICI evidence (DEV-CI-EVIDENCE-01)
QuestionsAsync questions
Peer contactDefault-off; peer collaboration
Composition/projection/JSNative execution
Browser proof/extraction/UI QABrowser routing
Library behaviorContext7 resolve-library-id → query-docs; otherwise official docs
Current/public/HTTP evidencecxc-search
Lost prior contextDEV-RECALL-01 (STRICT): cxc-recall chat+memory search; both miss: ask/report searches
Code smellscxc-dev-code-reviewer §3; boundary errors: dev-architecture §4
Types/static checksStatic analysis, gates
Lane observationWaiting (DISPATCH-POLL-BUDGET-01)
Capability gapInstalled skills come from the host's skill list and these routers. DEV-SKILL-DISCOVERY-01 (DEFAULT): only for gaps, the external catalog via cxc skill search/show; built-ins win
OwnershipOwnership; update owner first

Bundled files

The model reads these on demand while the skill is loaded. They are exposed as readable files and are never executed.

Frequently asked questions

What does the Cxc Dev AI skill do?

Use for coding, PR delivery, scaffolding and QA. Triggers: develop, fix, refactor, test, review, docs, browse, stacked PR, 개발, 수정, 검토, 스택 PR.

Why use Cxc Dev on TypingMind?

Because you install it once and use it with any model. Cxc Dev is plain Markdown rather than provider-specific code, so the same skill runs on GPT-5, Claude, Gemini, Grok, or a local model — and you can switch model mid-chat without it breaking. TypingMind runs on your own API keys, so you pay providers directly instead of a per-seat subscription, and your skills and chats stay in your own storage.

How do I install Cxc Dev in TypingMind?

Open Plugins → Skills → Install from GitHub in TypingMind and paste https://github.com/lidge-jun/codexclaw/tree/main/plugins/codexclaw/skills/dev. TypingMind reads its SKILL.md and bundles its files and installs it as a skill you can enable per chat.

Which AI models can use Cxc Dev?

Any model you connect in TypingMind. AI skills are plain Markdown instructions rather than provider-specific code, so GPT, Claude, Gemini, Grok, and local models can all load this skill when a request matches it.

How many AI models can I use with Cxc Dev?

As many as you like. As long as a model supports skills, you can use Cxc Dev with it — GPT, Claude, Gemini, Grok, DeepSeek, Mistral, Llama and more — all on TypingMind with your own API keys.

Is the Cxc Dev AI skill free?

Yes. It is published on GitHub by lidge-jun under the MIT license. You only pay your own AI provider for the tokens you use.

What are AI skills?

An AI skill is a reusable instruction bundle that teaches an AI model how to do one specific task. It follows the open Agent Skills format: a SKILL.md file with a name and description, plus any scripts, templates or reference files the model may need. The model reads the instructions only when your request matches the skill, so an installed skill costs nothing until it is used.

How are AI skills different from plugins or MCP servers?

A plugin or MCP server gives a model new tools to call — code that runs somewhere and returns a result. An AI skill gives the model knowledge and process instead: how to approach a task, which steps to follow, what good output looks like. Skills are plain Markdown, so they need no server, no API key and no runtime, and they work with any model.

View all

Set up your own AI workspace now

Get notified about new features and future giveaways by subscribing to our newsletter 👇