Fievel

View saved version →

---
name: ai-behavior-designer
description: Define AI responsibilities, permissions, uncertainty and recovery rules for a Fievel content system.
---

# AI Behavior Designer

## Using this skill from the Fievel plugin

This plugin carries Fievel's workflows and Python tools. The user's project is
separate from the installed plugin. Find the plugin root two directories above
this SKILL.md; its `runtime/` directory is the bundled resource root, called R below.
Resolve R from this skill's actual installed path, never a development checkout.

- Prefer Fievel's MCP tools when they are available (names start with `fievel_`).
  Local tools such as `fievel_init`, `fievel_validate`, `fievel_check` and
  `fievel_view` work only inside the user's workspace and never overwrite files;
  call `fievel_setup_status` first. Remote tools such as `fievel_validate_package`
  and `fievel_check_responses` receive package content, store nothing, and return
  results to save as new files; tell the user that content is sent to the Fievel
  server for processing. Without tools, use the CLI as described below.
- Read `R/AGENTS.md` for Fievel's conversation and workflow rules. References in
  this skill and in AGENTS.md to AGENTS.md, docs/, knowledge/, templates/, examples/,
  schemas/, scripts/ and .agents/skills/ are relative to R, not the user's working directory.
- Keep the current working directory in the user's chosen workspace. Run the CLI
  by its absolute path: `python "R/scripts/fievel.py" <command> <project> ...`.
  Replace R with the resolved path. Resolve project and output paths in the user's
  workspace. Run one simple command per step: the Python interpreter
  with the CLI path and its arguments, without chained commands, shell variables
  or PowerShell scripts. Read and list files with the host's file tools. Simple
  commands keep the user's permission prompts clear and predictable. Never create
  projects, install dependencies, or edit inside R or the plugin cache. Never copy Fievel AGENTS.md over a user's repository instructions.
- Check Python 3.10+ and `python -c "import jsonschema"`. If missing, explain the
  prerequisite. When setup is authorized, create a new `.fievel-venv` in the user's
  workspace and install `R/requirements.txt` with its Python. Reuse a working
  environment; do not overwrite an existing environment. Use that interpreter for
  every CLI command. Do not install globally or claim the plugin bundles Python.
- Follow existing workspace instructions. Explain that project files remain local,
  but material read into the conversation is handled by the user's coding agent.
- The repository-maintenance and adapter-sync instructions in R/AGENTS.md are for
  Fievel development, not for modifying the installed plugin during a user session.
- Other Fievel skills in this plugin: getting-started, content-system-builder,
  source-discovery, ai-behavior-designer, evaluation-designer and content-audit.
  Switch to the one that fits when the user's request changes.

Save projects under the user's chosen folder (default `my-content-system/` in
their workspace). The user approves content; validation does not. Do not run
model evaluations unless the user authorized them.

## Workflow

Read AGENTS.md and stage 4 of docs/WORKFLOW.md. Required inputs: product tasks,
actual available tools and the selected project's standards/sources. Ask what the
assistant should explain versus act on. Read knowledge/ai-experience-design/behavior.md
when explanation is useful. Define responsibility boundaries, clarification triggers,
confirmation for consequential actions, uncertainty, recovery and human handoff.

Write behavior.json with stable rule IDs and source links. Tool availability must
be verified; intended integrations are not working tools. The Wayfarer assistant
has no action tools, so must never claim to change or cancel bookings. Personalization
must use only provided context. Never infer hidden account information.

Output: draft behavior

Preview of the first 4,000 characters. Read full instructions →

View saved version →

---
name: content-audit
description: Audit an existing Fievel package for gaps, contradictory rules, source problems and evaluation coverage.
---

# Content Audit

## Using this skill from the Fievel plugin

This plugin carries Fievel's workflows and Python tools. The user's project is
separate from the installed plugin. Find the plugin root two directories above
this SKILL.md; its `runtime/` directory is the bundled resource root, called R below.
Resolve R from this skill's actual installed path, never a development checkout.

- Prefer Fievel's MCP tools when they are available (names start with `fievel_`).
  Local tools such as `fievel_init`, `fievel_validate`, `fievel_check` and
  `fievel_view` work only inside the user's workspace and never overwrite files;
  call `fievel_setup_status` first. Remote tools such as `fievel_validate_package`
  and `fievel_check_responses` receive package content, store nothing, and return
  results to save as new files; tell the user that content is sent to the Fievel
  server for processing. Without tools, use the CLI as described below.
- Read `R/AGENTS.md` for Fievel's conversation and workflow rules. References in
  this skill and in AGENTS.md to AGENTS.md, docs/, knowledge/, templates/, examples/,
  schemas/, scripts/ and .agents/skills/ are relative to R, not the user's working directory.
- Keep the current working directory in the user's chosen workspace. Run the CLI
  by its absolute path: `python "R/scripts/fievel.py" <command> <project> ...`.
  Replace R with the resolved path. Resolve project and output paths in the user's
  workspace. Run one simple command per step: the Python interpreter
  with the CLI path and its arguments, without chained commands, shell variables
  or PowerShell scripts. Read and list files with the host's file tools. Simple
  commands keep the user's permission prompts clear and predictable. Never create
  projects, install dependencies, or edit inside R or the plugin cache. Never copy Fievel AGENTS.md over a user's repository instructions.
- Check Python 3.10+ and `python -c "import jsonschema"`. If missing, explain the
  prerequisite. When setup is authorized, create a new `.fievel-venv` in the user's
  workspace and install `R/requirements.txt` with its Python. Reuse a working
  environment; do not overwrite an existing environment. Use that interpreter for
  every CLI command. Do not install globally or claim the plugin bundles Python.
- Follow existing workspace instructions. Explain that project files remain local,
  but material read into the conversation is handled by the user's coding agent.
- The repository-maintenance and adapter-sync instructions in R/AGENTS.md are for
  Fievel development, not for modifying the installed plugin during a user session.
- Other Fievel skills in this plugin: getting-started, content-system-builder,
  source-discovery, ai-behavior-designer, evaluation-designer and content-audit.
  Switch to the one that fits when the user's request changes.

Save projects under the user's chosen folder (default `my-content-system/` in
their workspace). The user approves content; validation does not. Do not run
model evaluations unless the user authorized them.

## Workflow

Read AGENTS.md. Required input: selected project folder and audit scope. Run
`python scripts/fievel.py validate <folder>` and `status` first. Uncovered draft
rules are reported as work remaining; run validate with --require-coverage when
assessing evaluation preparation, not ordinary early-stage draft integrity. Review source
authority/freshness, competing rules, approval status, missing context and ownership.
Check that scenarios cover actual tasks and failure situations, not just literal
phrases. Compare result system_sha256 with current package before drawing conclusions.

Output: a new audit Markdown file separating observed issues, uncertainties and
suggested changes, with rule/source IDs and supporting evidence. Do not edit rules
as part of a read-only audi

Preview of the first 4,000 characters. Read full instructions →

View saved version →

---
name: content-system-builder
description: Build or improve portable Fievel content standards and product context using user decisions and approved sources.
---

# Content System Builder

## Using this skill from the Fievel plugin

This plugin carries Fievel's workflows and Python tools. The user's project is
separate from the installed plugin. Find the plugin root two directories above
this SKILL.md; its `runtime/` directory is the bundled resource root, called R below.
Resolve R from this skill's actual installed path, never a development checkout.

- Prefer Fievel's MCP tools when they are available (names start with `fievel_`).
  Local tools such as `fievel_init`, `fievel_validate`, `fievel_check` and
  `fievel_view` work only inside the user's workspace and never overwrite files;
  call `fievel_setup_status` first. Remote tools such as `fievel_validate_package`
  and `fievel_check_responses` receive package content, store nothing, and return
  results to save as new files; tell the user that content is sent to the Fievel
  server for processing. Without tools, use the CLI as described below.
- Read `R/AGENTS.md` for Fievel's conversation and workflow rules. References in
  this skill and in AGENTS.md to AGENTS.md, docs/, knowledge/, templates/, examples/,
  schemas/, scripts/ and .agents/skills/ are relative to R, not the user's working directory.
- Keep the current working directory in the user's chosen workspace. Run the CLI
  by its absolute path: `python "R/scripts/fievel.py" <command> <project> ...`.
  Replace R with the resolved path. Resolve project and output paths in the user's
  workspace. Run one simple command per step: the Python interpreter
  with the CLI path and its arguments, without chained commands, shell variables
  or PowerShell scripts. Read and list files with the host's file tools. Simple
  commands keep the user's permission prompts clear and predictable. Never create
  projects, install dependencies, or edit inside R or the plugin cache. Never copy Fievel AGENTS.md over a user's repository instructions.
- Check Python 3.10+ and `python -c "import jsonschema"`. If missing, explain the
  prerequisite. When setup is authorized, create a new `.fievel-venv` in the user's
  workspace and install `R/requirements.txt` with its Python. Reuse a working
  environment; do not overwrite an existing environment. Use that interpreter for
  every CLI command. Do not install globally or claim the plugin bundles Python.
- Follow existing workspace instructions. Explain that project files remain local,
  but material read into the conversation is handled by the user's coding agent.
- The repository-maintenance and adapter-sync instructions in R/AGENTS.md are for
  Fievel development, not for modifying the installed plugin during a user session.
- Other Fievel skills in this plugin: getting-started, content-system-builder,
  source-discovery, ai-behavior-designer, evaluation-designer and content-audit.
  Switch to the one that fits when the user's request changes.

Save projects under the user's chosen folder (default `my-content-system/` in
their workspace). The user approves content; validation does not. Do not run
model evaluations unless the user authorized them.

## Workflow

Read AGENTS.md, docs/WORKFLOW.md and the selected project's progress.json.
Required inputs: selected project folder, product/audience intent, and any supplied
standards. Start with stages 1 and 2; ask only for missing decisions. When the
user supplies or names a source document, run the source-discovery skill first to
inventory and assess it; then use the existing-guidelines path to map its rules,
preserving their source IDs and approval status. Do not turn inferred preferences into approved organization rules.

For existing material, read knowledge/content-design/using-existing-guidelines.md.
Show a small source-to-proposed-rule mapping with precise evidence location where
available. Preserve meaning and distinguish explicit requiremen

Preview of the first 4,000 characters. Read full instructions →

View saved version →

---
name: evaluation-designer
description: Create plain-language Fievel quality checks, check saved responses, or prepare human review without claiming unexecuted tests ran.
---

# Evaluation Designer

## Using this skill from the Fievel plugin

This plugin carries Fievel's workflows and Python tools. The user's project is
separate from the installed plugin. Find the plugin root two directories above
this SKILL.md; its `runtime/` directory is the bundled resource root, called R below.
Resolve R from this skill's actual installed path, never a development checkout.

- Prefer Fievel's MCP tools when they are available (names start with `fievel_`).
  Local tools such as `fievel_init`, `fievel_validate`, `fievel_check` and
  `fievel_view` work only inside the user's workspace and never overwrite files;
  call `fievel_setup_status` first. Remote tools such as `fievel_validate_package`
  and `fievel_check_responses` receive package content, store nothing, and return
  results to save as new files; tell the user that content is sent to the Fievel
  server for processing. Without tools, use the CLI as described below.
- Read `R/AGENTS.md` for Fievel's conversation and workflow rules. References in
  this skill and in AGENTS.md to AGENTS.md, docs/, knowledge/, templates/, examples/,
  schemas/, scripts/ and .agents/skills/ are relative to R, not the user's working directory.
- Keep the current working directory in the user's chosen workspace. Run the CLI
  by its absolute path: `python "R/scripts/fievel.py" <command> <project> ...`.
  Replace R with the resolved path. Resolve project and output paths in the user's
  workspace. Run one simple command per step: the Python interpreter
  with the CLI path and its arguments, without chained commands, shell variables
  or PowerShell scripts. Read and list files with the host's file tools. Simple
  commands keep the user's permission prompts clear and predictable. Never create
  projects, install dependencies, or edit inside R or the plugin cache. Never copy Fievel AGENTS.md over a user's repository instructions.
- Check Python 3.10+ and `python -c "import jsonschema"`. If missing, explain the
  prerequisite. When setup is authorized, create a new `.fievel-venv` in the user's
  workspace and install `R/requirements.txt` with its Python. Reuse a working
  environment; do not overwrite an existing environment. Use that interpreter for
  every CLI command. Do not install globally or claim the plugin bundles Python.
- Follow existing workspace instructions. Explain that project files remain local,
  but material read into the conversation is handled by the user's coding agent.
- The repository-maintenance and adapter-sync instructions in R/AGENTS.md are for
  Fievel development, not for modifying the installed plugin during a user session.
- Other Fievel skills in this plugin: getting-started, content-system-builder,
  source-discovery, ai-behavior-designer, evaluation-designer and content-audit.
  Switch to the one that fits when the user's request changes.

Save projects under the user's chosen folder (default `my-content-system/` in
their workspace). The user approves content; validation does not. Do not run
model evaluations unless the user authorized them.

## Workflow

Read AGENTS.md and stages 5–6 of docs/WORKFLOW.md. Required inputs: actual design
requirements, realistic user situations and a project folder. Read
knowledge/evaluations/checking-quality.md on demand. Translate “good” into inspectable
questions; link each check's requirement_id to a standards or behavior rule.
Cover ordinary, unknown, conflicting-source and action-request situations.

Use contains/not_contains only for case-insensitive literal checks. Use human for
meaning, invented facts, tone, accessibility and action authorization. A phrase
absence cannot prove the absence of unauthorized action. Store definitions in
evaluation.json. Drafts can be saved before coverage is complete; before execution,
run `python scripts/fi

Preview of the first 4,000 characters. Read full instructions →

View saved version →

---
name: getting-started
description: Start or resume Fievel onboarding and choose a beginner, existing-guidelines, or AI-experience workflow.
---

# Getting Started

## Using this skill from the Fievel plugin

This plugin carries Fievel's workflows and Python tools. The user's project is
separate from the installed plugin. Find the plugin root two directories above
this SKILL.md; its `runtime/` directory is the bundled resource root, called R below.
Resolve R from this skill's actual installed path, never a development checkout.

- Prefer Fievel's MCP tools when they are available (names start with `fievel_`).
  Local tools such as `fievel_init`, `fievel_validate`, `fievel_check` and
  `fievel_view` work only inside the user's workspace and never overwrite files;
  call `fievel_setup_status` first. Remote tools such as `fievel_validate_package`
  and `fievel_check_responses` receive package content, store nothing, and return
  results to save as new files; tell the user that content is sent to the Fievel
  server for processing. Without tools, use the CLI as described below.
- Read `R/AGENTS.md` for Fievel's conversation and workflow rules. References in
  this skill and in AGENTS.md to AGENTS.md, docs/, knowledge/, templates/, examples/,
  schemas/, scripts/ and .agents/skills/ are relative to R, not the user's working directory.
- Keep the current working directory in the user's chosen workspace. Run the CLI
  by its absolute path: `python "R/scripts/fievel.py" <command> <project> ...`.
  Replace R with the resolved path. Resolve project and output paths in the user's
  workspace. Run one simple command per step: the Python interpreter
  with the CLI path and its arguments, without chained commands, shell variables
  or PowerShell scripts. Read and list files with the host's file tools. Simple
  commands keep the user's permission prompts clear and predictable. Never create
  projects, install dependencies, or edit inside R or the plugin cache. Never copy Fievel AGENTS.md over a user's repository instructions.
- Check Python 3.10+ and `python -c "import jsonschema"`. If missing, explain the
  prerequisite. When setup is authorized, create a new `.fievel-venv` in the user's
  workspace and install `R/requirements.txt` with its Python. Reuse a working
  environment; do not overwrite an existing environment. Use that interpreter for
  every CLI command. Do not install globally or claim the plugin bundles Python.
- Follow existing workspace instructions. Explain that project files remain local,
  but material read into the conversation is handled by the user's coding agent.
- The repository-maintenance and adapter-sync instructions in R/AGENTS.md are for
  Fievel development, not for modifying the installed plugin during a user session.
- Other Fievel skills in this plugin: getting-started, content-system-builder,
  source-discovery, ai-behavior-designer, evaluation-designer and content-audit.
  Switch to the one that fits when the user's request changes.

Save projects under the user's chosen folder (default `my-content-system/` in
their workspace). The user approves content; validation does not. Do not run
model evaluations unless the user authorized them.

## Workflow

Read AGENTS.md and docs/WORKFLOW.md. If a project exists, read its progress.json
and validate it before asking for information again. Briefly summarize the saved
decisions, open questions, and next stage. Do not re-onboard an experienced user.

For a new user, introduce Fievel as an AI content systems architect, then offer:
“I’m new to this”, “I have existing content guidelines”, or “I’m building an AI
experience”. Ask which path fits. Explain that a content system connects product
context, rules, sources, behavior, checks, and ownership. Offer the Wayfarer example
without requiring company data. Ask one useful question at a time and say why.

Use knowledge/INDEX.md to load one relevant local explanation; fundamental concepts
and the fictional exercise should

Preview of the first 4,000 characters. Read full instructions →

View saved version →

---
name: source-discovery
description: Inventory and assess local or approved sources for a Fievel content system, preserving authority and provenance. Use when a user brings a style guide, policy or other document into their system, before its rules are mapped.
---

# Source Discovery

## Using this skill from the Fievel plugin

This plugin carries Fievel's workflows and Python tools. The user's project is
separate from the installed plugin. Find the plugin root two directories above
this SKILL.md; its `runtime/` directory is the bundled resource root, called R below.
Resolve R from this skill's actual installed path, never a development checkout.

- Prefer Fievel's MCP tools when they are available (names start with `fievel_`).
  Local tools such as `fievel_init`, `fievel_validate`, `fievel_check` and
  `fievel_view` work only inside the user's workspace and never overwrite files;
  call `fievel_setup_status` first. Remote tools such as `fievel_validate_package`
  and `fievel_check_responses` receive package content, store nothing, and return
  results to save as new files; tell the user that content is sent to the Fievel
  server for processing. Without tools, use the CLI as described below.
- Read `R/AGENTS.md` for Fievel's conversation and workflow rules. References in
  this skill and in AGENTS.md to AGENTS.md, docs/, knowledge/, templates/, examples/,
  schemas/, scripts/ and .agents/skills/ are relative to R, not the user's working directory.
- Keep the current working directory in the user's chosen workspace. Run the CLI
  by its absolute path: `python "R/scripts/fievel.py" <command> <project> ...`.
  Replace R with the resolved path. Resolve project and output paths in the user's
  workspace. Run one simple command per step: the Python interpreter
  with the CLI path and its arguments, without chained commands, shell variables
  or PowerShell scripts. Read and list files with the host's file tools. Simple
  commands keep the user's permission prompts clear and predictable. Never create
  projects, install dependencies, or edit inside R or the plugin cache. Never copy Fievel AGENTS.md over a user's repository instructions.
- Check Python 3.10+ and `python -c "import jsonschema"`. If missing, explain the
  prerequisite. When setup is authorized, create a new `.fievel-venv` in the user's
  workspace and install `R/requirements.txt` with its Python. Reuse a working
  environment; do not overwrite an existing environment. Use that interpreter for
  every CLI command. Do not install globally or claim the plugin bundles Python.
- Follow existing workspace instructions. Explain that project files remain local,
  but material read into the conversation is handled by the user's coding agent.
- The repository-maintenance and adapter-sync instructions in R/AGENTS.md are for
  Fievel development, not for modifying the installed plugin during a user session.
- Other Fievel skills in this plugin: getting-started, content-system-builder,
  source-discovery, ai-behavior-designer, evaluation-designer and content-audit.
  Switch to the one that fits when the user's request changes.

Save projects under the user's chosen folder (default `my-content-system/` in
their workspace). The user approves content; validation does not. Do not run
model evaluations unless the user authorized them.

## Workflow

Read AGENTS.md and stage 3 of docs/WORKFLOW.md. Required input: selected project
and the local locations or connections the user wants assessed. Ask where standards
are maintained and who owns them. Inventory before ingesting. Record title,
location, owner (null if unknown), purpose, authority, review date (null if unknown),
permissions, permitted use and ingested in sources.json.

Use knowledge/content-design/using-existing-guidelines.md when converting supplied
standards, and knowledge/content-design/system-relationships.md when explaining
their connections. Bundled lessons teach the method; they cannot substitute for
the user's organizat

Preview of the first 4,000 characters. Read full instructions →