The Agent Skills Directory

[PROMPT_INJECTION]: The skill instructs the agent to disable platform-level acceptance gates for its analysis sub-agents in SKILL.md ('Where the worker mechanism offers acceptance gates, disable them'). This is designed to ensure the agent can process and report on code containing security vulnerabilities, though it involves a bypass of standard platform safety filters.
[COMMAND_EXECUTION]: The skill executes local shell commands to audit the target repository, including bin/rails routes, bundle exec brakeman, bundle audit, and various grep-based searches across the references/lenses/ rubrics. These tools are used for their intended purpose of mapping and auditing Rails application security.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze untrusted Rails application code, creating an attack surface for indirect prompt injection.
Ingestion points: Source code and configuration files are read from the app/, lib/, config/, and db/ directories as defined in SKILL.md.
Boundary markers: The skill implements a 'blind agent' architecture in SKILL.md where each analysis lens is executed in a fresh context, providing a structural boundary.
Capability inventory: The skill uses local shell commands (grep, rails, brakeman, bundle audit, npm audit) across all lens files in references/lenses/ to inspect the repository content.
Sanitization: No explicit sanitization of the source code content is performed before it is processed by the analysis lenses.
[EXTERNAL_DOWNLOADS]: The skill performs network operations to update vulnerability databases using standard tools such as bundle audit check --update and npm audit in references/lenses/dependencies.md. These operations target official package registries and well-known vulnerability databases.

rails-security