code-review

title: .mustflow/skills/*/SKILL.md description: A procedural document defining repeatable tasks for agents.

.mustflow/skills/*/SKILL.md defines specific procedures to help agents execute repeatable tasks without inference or guesswork.

Usage

Agents identify the appropriate skill from .mustflow/skills/INDEX.md and consult its documentation before performing repeatable tasks.

Skill documents encompass procedures such as code review, test maintenance, failure triage, and documentation updates. They reference the global policies defined in .mustflow/docs/agent-workflow.md rather than duplicating them.

Activating a skill means reading and following the skill procedure. It does not grant permission to run commands outside .mustflow/config/commands.toml or to ignore higher-priority instructions.

Frontmatter

mustflow_doc: skill.code-review
locale: en
canonical: true
revision: 1
name: code-review
description: Apply this skill when reviewing code changes, scope, risks, or verification gaps.
metadata:
  mustflow_schema: "1"
  mustflow_kind: procedure
  pack_id: mustflow.core
  skill_id: mustflow.core.code-review
  command_intents:
    - test
    - lint

• mustflow_doc: A unique, stable identifier for the skill within mustflow.
• locale: The language code of the document.
• canonical: Indicates if the document serves as the authoritative source.
• revision: The revision number of the authoritative document.
• name: The skill name. It must match the containing .mustflow/skills/<name>/ directory.
• description: A brief summary of when the agent should apply this skill.
• metadata.mustflow_schema: The version of the skill metadata schema. The current supported value is "1".
• metadata.mustflow_kind: The category of the document. Default skills must use procedure.
• metadata.pack_id: The package or pack namespace that owns the skill, such as mustflow.core.
• metadata.skill_id: The globally scoped skill identifier. It must combine the pack identifier and folder name, such as mustflow.core.code-review.
• metadata.command_intents: A list of command intent names referenced by this skill. Each name must exist in .mustflow/config/commands.toml.

The English version of the skill template serves as the authoritative source. Localized versions specify their own locales and set canonical: false.

Document Structure

Each skill document should include the following sections:

• Purpose: Defines the specific task or objective this skill addresses.
• Use When: Describes scenarios that should trigger the application of this skill.
• Do Not Use When: Specifies exclusions to prevent unnecessary application.
• Required Inputs: Context and information the agent must collect before execution.
• Preconditions: Conditions that must hold before following the procedure.
• Allowed Edits: The files or surfaces the skill may guide, without granting command permission.
• Procedure: The step-by-step sequence of operations.
• Postconditions: Evidence, outputs, and unresolved risks that must be reportable after the procedure.
• Verification: Relevant command intents and verification steps.
• Failure Handling: Protocols for managing command failures or missing information.
• Output Format: Required components to include in the final report.

Authoring Guidelines

Each skill should be scoped to a single task type.

Avoid including raw shell commands within skill documents. In the Verification section, reference the .mustflow/docs/agent-workflow.md#command-execution-policy and list only the relevant command intent names.

Each intent must be resolved via .mustflow/config/commands.toml. If an intent is not marked as status = "configured", it must not be executed; instead, report its status and the reason for skipping.

Do not write that a skill itself grants command execution permission. Skills describe procedure; .mustflow/config/commands.toml remains the only source for runnable command permissions.

Example:

## Verification

Relevant command intents:

- `test`
- `lint`

Resolve each intent through `.mustflow/config/commands.toml`.

Resource Management

A default skill begins with only a SKILL.md file. Do not create empty references/, assets/, or scripts/ directories in advance.

If a skill requires extensive supporting material, an optional resources.toml file can be added to register references, assets, or scripts. Scripts should never be invoked via inferred paths; they must be mapped to specific command intents in .mustflow/config/commands.toml.

Refer to Skill Resources for detailed policies and guidelines.