service-knowledge-contour

Knowledge Contour Skills

Language: English | Русский

This repository contains ready-to-install Codex skills: self-contained folders with SKILL.md, agent prompt files, references, scripts, and tests.

Skills

product-workflow

End-to-end product shaping: discovery, decision log, PRD, user scenarios, current-scenario baseline, feature impact analysis, implementation plan, hardening plan, independent verification, editorial pass, and stakeholder-facing PDF.

Use it for product or feature descriptions, roadmap validation, scenario design, and implementation planning. For existing products, the baseline mode captures current scenarios as current-scenario-baseline.md, scenario-cards.md, and scenario-graph.dot; new increments then declare affected scenarios through pre-scan and impact artifacts before implementation planning. By default, the PDF includes only the product problem, scenarios, chosen solution, and independent verdict; T/H plans remain engineering artifacts.

flowchart TD
    U["User vision"] --> D["Phase 0: discovery agents"]
    D --> DL["Decision log: proposed / approved / delegated"]
    DL --> H{"Human approval?"}
    H -->|approved| S["Scenarios + overview"]
    H -->|needs choice| U
    S --> B["Baseline: scenario cards + DOT graph"]
    B --> I["Increment pre-scan + impact"]
    I --> C["Scenario critic loop"]
    C --> P["Implementation plan with Product artifacts"]
    P --> HP["Hardening plan"]
    HP --> V["Independent artifact verifier"]
    V -->|blockers| C
    V -->|approved verdict| E["Editorial style pass"]
    E --> PDF["Product PDF"]
    P --> IV["Post-implementation verifier"]
    HP --> IV

Contents:

• SKILL.md — main workflow and gates.
• agents/ — discovery, critic, verifier, and style-editor prompts.
• references/ — templates for scenarios, baseline, scenario cards, increment pre-scan/impact, decision log, plans, and PDF.
• scripts/verify_artifacts.py — structural checks for scenarios, baseline/pre-scan/impact artifacts, plans, hardening plans, and validation gate.
• scripts/build_pdf.sh — PDF assembly with mandatory independent validation.
• evals/ — expected-behavior eval set.

service-knowledge-contour

Minimal knowledge contour for one service repository: startup docs, canonical SERVICE_MAP.md / VERIFY.md, knowledge-gap registry, generated overlays, audit, promotion, and pruning.

Use it when a service needs a stable operating knowledge layer for humans and agents, onboarding docs are missing or fragmented, or topology, entrypoints, verification commands, integrations, or risk zones changed.

flowchart TD
    R["Repository reality"] --> B["bootstrap"]
    B --> C["Canonical core"]
    C --> G["Generated layer"]
    G --> A["audit / strict audit"]
    A -->|trigger fired| H{"Human approval needed?"}
    H -->|yes| P["promote / repair candidate"]
    H -->|no| G
    P --> C
    C --> V["Independent contour verifier"]
    A --> PR["PR / CI evidence"]

Contents:

• SKILL.md — service knowledge contour workflow and rules.
• agents/contour-verifier.md — independent semantic contour verification.
• bin/ — bootstrap, refresh, audit, promote, and prune shell scripts.
• examples/ — GitHub Actions and PR template examples.
• tests/ — bootstrap/audit contract tests.

Install In Codex

Install a skill by pointing Codex at the repository folder URL.

Ask Codex:

Install the skill from https://github.com/ehlyzov/skills/tree/main/product-workflow

or:

Install the skill from https://github.com/ehlyzov/skills/tree/main/service-knowledge-contour

Manual install:

mkdir -p ~/.codex/skills
cp -R product-workflow ~/.codex/skills/
cp -R service-knowledge-contour ~/.codex/skills/

Update an installed copy:

rm -rf ~/.codex/skills/product-workflow ~/.codex/skills/service-knowledge-contour
cp -R product-workflow service-knowledge-contour ~/.codex/skills/

Verify installation:

test -f ~/.codex/skills/product-workflow/SKILL.md
test -f ~/.codex/skills/service-knowledge-contour/SKILL.md

Install In Claude Code

Add this repository as a Claude Code plugin marketplace:

/plugin marketplace add ehlyzov/skills

or, with the full Git URL:

/plugin marketplace add https://github.com/ehlyzov/skills.git

Then install one or both plugins from the marketplace:

/plugin install product-workflow@knowledge-contour-skills
/plugin install service-knowledge-contour@knowledge-contour-skills

The marketplace file is stored at .claude-plugin/marketplace.json, which is the path Claude Code expects when adding a GitHub repository as a plugin marketplace.

Run Scripts

Running scripts is separate from installing a skill.

For product-workflow, run scripts from the skill folder or by absolute path to the installed skill: