lov-skill-publisher

lov-skill-publisher

Publish one validated local Skill source to the LovStudio official website or explicitly selected additional channels. When the user does not specify a channel, run only the Skill Publisher website adapter. Always auto-price first, keep channel metadata and generated packages outside canonical source, and report evidence per selected channel.

Triggers

Activate when

• The user says "publish this Skill", "list it on Skill Publisher", "generate a WorkBuddy package", or "distribute to multiple platforms".
• The user asks to publish, release, distribute, upload, or package an existing Skill.

Do not activate when

• The user wants to create, implement, modify, or only install a Skill locally; hand it to lov-skill-creator。
• The user is merely invoking a certain business Skill, not publishing it.

Product boundary

• Input is a local Skill source that already passes source validation.
• When no channel is specified, select only Skill Publisher, the LovStudio official website adapter.
• Explicitly named channels narrow the run; do not ask a channel-selection question when the request omits channel parameters.
• Run every supported adapter only when the user explicitly says "全部渠道", "多平台", "all channels", or names every intended channel.
• A request may select multiple channels in one run.
• Pricing, visibility, protection, licensing, and target accounts are publishing inputs. Reuse context when known and ask only for values required by a target.
• A license scope named global or all is a dynamic entitlement to every currently listed Skill for the license lifetime. Never publish or migrate it as a frozen list of current Skill IDs; explicit per-Skill grants remain fixed.
• Every publish run invokes lov-skill-pricing by default to create or refresh one evidence-backed Pricing Card per Skill before channel preparation. An explicit user price is a constraint for that pricing pass, not a reason to skip it.
• Keep the canonical Pricing Card in publisher profile/output storage. Transform only its public fields into catalogs, packages, or submission forms; do not add channel state or generated cards to canonical Skill source.
• Channel metadata, credentials, staging files, and archives stay outside source.

Supported adapters in this version:

• Skill Publisher — source repository, release, catalog, cache refresh, and live page.
• WorkBuddy (CodeBuddy Open Platform) — Generate a standalone Skill ZIP, and upload, parse, fill in listing information, and submit for review on the console.
• Alipay SkillPay — validated product ZIP, explicit CNY price, upload, parse, submission, and observable review state.

For any additional platform, follow references/channels.md and verify its current official name, submission contract, public URL, and completion signal before implementing an adapter.

User Configuration

Publishing inherently needs persistent target settings. Initialize them on first use through references/user-config.md, while keeping tokens in environment or credential stores rather than committed profiles.

Workflow (MANDATORY)

Step 0: Resolve roots and settings

• Resolve this Skill as SKILL_DIR.
• Resolve the source from an explicit path, current directory, or conversation.
• Resolve target settings from flags, environment, and shared profile.
• Verify referenced scripts and channel documents before external changes.

Step 1: Validate canonical source

python3 "$SKILL_DIR/scripts/validate_skill.py" SOURCE --target source

Check that source has no platform metadata directory or generated release artifacts. Record name, version, description, modules, dependencies, git state, and whether a remote already exists.

Step 2: Auto-price the Skill

Use lov-skill-pricing for every source, even when the user only says "发布" or selects a channel without mentioning price.

1. Pass the validated source, version, delivery unit, real result evidence, maintenance/support conditions, selected channels, and any existing price or pricing-card.yaml into the pricing workflow.
2. Let the pricing workflow use explicit assumptions when cost/value inputs are missing. Ask at most one focused question only when the missing field changes the commercial model; ordinary missing estimates must not block publication.
3. If the user supplied a public price, currency, free/paid status, or billing model, preserve it as a hard publishing constraint and have the Pricing Card explain any difference from its model recommendation.
4. Record recommended price, launch price, stable range, billing model, channel, cost floor, value anchor, weighted score, confidence, evidence gaps, and review trigger outside canonical source.
5. Reuse this one price contract across selected adapters. Skill Publisher maps it into catalog pricing metadata and its public Pricing Card; SkillPay uses the public CNY price; installation-only channels use the free/paid funnel and upgrade path without inventing a separate price.

Do not hand-author an unexplained price inside a channel adapter. If automatic pricing cannot run, mark pricing blocked with the missing Skill/resource and do not submit a paid listing with an inferred number.

Step 3: Resolve channels and release model

If channels are explicit, proceed without another distribution question. If no channel is named, select only Skill Publisher and proceed without asking the user to choose a channel. Expand to all supported adapters only after an explicit all-channel or multi-platform request.

For each selected channel, resolve only required fields:

• public/private visibility and free/paid catalog status where supported;
• paid delivery mode: protected encrypted bundle or explicitly public source;
• organization, account, catalog, or output location;
• platform metadata, icon, examples, and source locator;
• requested version versus current source version.

Do not ask users to choose implementation details such as staging layout, validation commands, archive format, or adapter order.

Step 4: Build a per-channel plan

Read references/channels.md, then load only the selected channel references. Keep independent state for each target so one failure does not masquerade as a successful multi-channel release.

Step 5: Publish Skill Publisher

Read references/publishing.md completely. Execute the source repository, release, catalog, cache refresh, and live verification workflow. Publication is complete only when the expected version and release-specific content are visible on the live detail page and the catalog's exact install command succeeds through the declared delivery mode. A paid catalog entry without either a verified encrypted bundle or explicit public_source: true is blocked, not published. When a catalog refresh adds or delists a Skill, verify that dynamic global/all licenses immediately gain or lose catalog access without granting or spending Credits. This entitlement check is part of publication, not a later migration.

Step 6: Publish WorkBuddy through CodeBuddy

Read references/workbuddy.md completely. Keep connector metadata and icon in a publisher profile outside source, then run:

python3 "$SKILL_DIR/scripts/build_workbuddy.py" SOURCE \
  --meta CONNECTOR_META \
  --icon ICON \
  --output-dir OUTPUT_DIR

Record source validation, package validation, archive listing, checksum, module count, and output paths. Then follow references/workbuddy.md: upload each individual ZIP to https://www.codebuddy.cn/open/console/dashboard, wait for "解析成功", fill user-facing Chinese/English listing fields, and submit it. Record the review state separately from public listing evidence.

Step 7: Submit Alipay SkillPay

Read references/skillpay.md completely. Build a clean product ZIP outside the canonical source, keep the current Pricing Card price as the single public CNY price, then upload and wait for parsing to finish before submitting. Record the product title, package checksum, public price, submission result, and current review state. A parsed archive is only uploaded; a success notice after form submission is review until the marketplace marks the product live.

Step 8: Additional platform adapter

Use the adapter contract in references/channels.md. Research current official documentation, implement deterministic preparation/validation scripts when useful, and define an observable completion gate. Never reuse another channel's metadata or call an upload dialog a completed publication.

Step 9: Multi-channel report

Report each target separately:

Channel	State	Version/artifact	Evidence	Follow-up
TARGET	prepared/published/verified	VALUE	URL or local path	ACTION

Use precise states. prepared, uploaded, installed, listed, and live represent different outcomes.

Dependencies

• Python 3.8+
• PyYAML
• lov-skill-pricing for the default pre-publish Pricing Card
• git and gh for GitHub-backed publication
• Target-specific credentials resolved without printing secrets

Local development

Validate this publisher Skill with:

python3 scripts/validate_skill.py . --target source

Runtime context (shared)

Before running, read this Skill package's skill.yaml, with the host providing the skill-runtime/v1 context. The field resolution order is: current request, project context, personal Preferences, brand Profile, general defaults.

• Use only the fields declared in the Manifest; Profile stores public brand facts, and Preferences stores personal work preferences.
• required: true When a field is missing, ask the user one focused question according to the Manifest's question configuration; save the answer only after the user explicitly agrees.
• Errors provide a copyable context_id, field path, and source; diagnostic content avoids secrets, full private paths, and raw configuration.

General feedback loop

When the user raises modification feedback during a Skill-driven task, the following must be executed before continuing with the current artifact:

1. First determine whether the feedback is task-specific(this time only) or reusable(reusable across tasks).
2. task-specific Modify only the current task, not the Skill.
3. reusable First determine the scope: for domain rules, first update the corresponding canonical Skill; for rules applicable to all Skills, first update the shared specification.
4. After completing rule updates, versioning, lint, and distribution verification, then apply the modifications to the current task.
5. reusable Modifications invalidate previous "confirm", "continue", and "go ahead"; after completing the current artifact modification and read-back, you must stop and wait for the user's next instruction, without automatically proceeding to publishing, submission, or other external writes.