Vic-distill-industry-skills

Vic Agent Skills

简体中文

An open-source collection of self-contained, testable, and portable Agent Skills for industry research, business monetization, and professional workflows.

Available skills

distill-industry

Research and distill an industry into an evidence-backed knowledge pack, opportunity map, entry playbook, capability map, and tested atomic Agent Skills.

• Supports scan, full, skill-pack, refresh, compare, and audit modes.
• Separates time-sensitive market evidence from durable methods and workflows.
• Runs an 11-stage resumable pipeline with V1 evidence, V2 decision, and V3 execution gates.
• Maps the value chain, roles, lifecycle, and recurring tasks before selecting atomic Skills.
• Preserves every candidate as accepted, deferred, or rejected with an auditable reason.
• Supports one claim with multiple independent source relationships and explicit conflict handling.
• Includes deterministic initialization, validation, formula regression, result merging, and orchestration scripts.
• Provides isolated multi-agent execution with a sequential fallback when workers are unavailable.

The current stable release is distill-industry v2.0.0. Stable describes the meta Skill's workflow, schemas, validators, and fallback behavior. Every generated industry pack still receives its own draft, pilot, or production readiness grade.

build-monetization-loop

Turn a selected opportunity, expertise, product, service, or underperforming business into a measurable end-to-end monetization system.

• Supports diagnose, design, validate, launch, optimize, scale, and audit modes.
• Covers customer evidence, offers, pricing, unit economics, acquisition, sales, payment, delivery, retention, referral, and automation.
• Uses a 12-stage resumable workflow and five commercial validation gates.
• Reconciles funnel activity with collected cash, delivery cost, quality, refunds, and capacity.
• Includes 15 canonical operating templates, deterministic validators, formula regression tests, and behavioral test cases.
• Runs independently. An industry-research pack can be used as optional input but is not a dependency.

Repository structure

Vic-Agents-Skills/
├── README.md
├── README.zh-CN.md
├── LICENSE
└── skills/
    ├── distill-industry/
    │   ├── SKILL.md
    │   ├── agents/
    │   ├── assets/
    │   ├── examples/
    │   ├── references/
    │   ├── scripts/
    │   └── tests.json
    └── build-monetization-loop/
        ├── SKILL.md
        ├── agents/
        ├── assets/
        ├── examples/
        ├── references/
        ├── scripts/
        └── tests.json

Every directory under skills/ is a self-contained installable Skill. Repository documentation stays outside Skill directories so it does not consume an agent's runtime context.

Installation

1. Clone or download this repository.
2. Copy the complete directory of the Skill you need into the skills directory supported by your Agent platform.
3. Keep SKILL.md and all bundled resources together.
4. Invoke the Skill by name on a concrete request.

Example for Claude Code on macOS or Linux:

git clone https://github.com/shangdizhiyan/Vic-Agents-Skills.git
mkdir -p ~/.claude/skills
cp -R Vic-Agents-Skills/skills/build-monetization-loop ~/.claude/skills/

For ChatGPT, Codex, or another Agent Skills implementation, use that platform's current installation flow. Platform-specific tool permissions and UI metadata may differ even when the portable core is readable.

Example requests

• Use distill-industry to compare transformer export opportunities in the UAE, Malaysia, and Indonesia using one evidence standard.
• Use build-monetization-loop to turn a validated web-development capability into an offer, acquisition, sales, delivery, retention, and automation system.
• Diagnose why a business has booked revenue but weak cash flow and uncertain contribution margin.
• Determine which CRM, follow-up, pricing, and retention tasks can be automated after manual sales and delivery have been validated.

More examples are bundled in each Skill's examples/sample-requests.md.

Validation

Validate a generated industry pack:

python3 skills/distill-industry/scripts/validate_pack.py /path/to/industry-pack
python3 skills/distill-industry/scripts/run_formula_tests.py /path/to/industry-pack
python3 skills/distill-industry/scripts/validate_parallel_run.py /path/to/industry-pack

Initialize and validate a monetization pack:

python3 skills/build-monetization-loop/scripts/init_loop.py /path/to/monetization-pack --business "Example Business" --mode design
python3 skills/build-monetization-loop/scripts/validate_loop.py /path/to/monetization-pack
python3 skills/build-monetization-loop/scripts/run_economics_tests.py /path/to/monetization-pack

The repository is checked for valid Agent Skills frontmatter, complete local resource references, parseable Python/JSON/YAML, repeatable initialization without overwrite, formula regressions, behavioral trigger cases, safety boundaries, and absence of credentials or build artifacts. Static validation does not prove real demand, profit, retention, or production readiness.

Design lineage

This project takes inspiration from the open-source Cangjie Skill, particularly its emphasis on atomic boundaries, progressive disclosure, resumable pipelines, validation, audit trails, and trigger-confusion testing. It does not copy Cangjie's book-distillation workflow or RIA-TV++ templates. The monetization Skill uses an original business operating model centered on customer behavior, unit economics, cash collection, fulfillment, retention, and controlled automation.

Compatibility

The portable core uses only name and description frontmatter and avoids requiring platform-only syntax. Optional agents/openai.yaml files provide OpenAI-facing UI metadata. Compatibility depends on each host's support for file-based Skills, tools, networking, subagents, and filesystem writes; identical behavior across all Agents is not guaranteed.

Contributing

Issues and pull requests are welcome. Include a reproducible request, observed behavior, expected behavior, and relevant validation output. Keep dynamic facts out of durable Skill instructions unless the Skill also defines source, region, valid date, and refresh rules.

License

Released under the MIT License.