The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes local shell commands via the Salesforce CLI (sf) to detect package namespaces and retrieve Salesforce object metadata. These operations are performed using execFileSync in scripts/namespace.mjs and described in references/get-object-documentation.md. The commands are used to inspect the local environment and the connected Salesforce org.
[INDIRECT_PROMPT_INJECTION]: The skill ingests and processes data from local content.json files that define the configuration of Salesforce Digital Experience site views.
Ingestion points: Files located at force-app/main/default/digitalExperiences/site/<site>/sfdc_cms__view/<view>/content.json are read to identify existing component structures.
Boundary markers: The skill relies on strict JSON parsing and structural walking of the component tree but does not implement specific delimiters to separate data from instructions within the JSON attributes.
Capability inventory: The skill has the capability to write to the local file system and execute shell commands through the Salesforce CLI.
Sanitization: The skill employs a validation script (scripts/validate-view.mjs) to verify the structural integrity of modified files and a regex-based whitelist (SAFE_SF_ARG_RE) in scripts/namespace.mjs to sanitize arguments before shell execution.
[SAFE]: The vendor-specific resources and namespaces (avxp, avcmpbuilder) align with the documented author profile ('avonni'). Additionally, the skill implements defensive programming by sanitizing shell arguments to prevent command injection when operating on Windows systems where shell: true is required for .cmd files.

avonni-experience-components