The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it ingests untrusted code from the local directory and uses sub-agents with write permissions to process it.
Ingestion points: In Phase 4 (Interactive Generation), the sub-agents read function/class/type implementations from the codebase to understand their behavior.
Boundary markers: The instructions for sub-agents lack explicit boundary markers or directions to ignore natural language instructions found within the code comments or string literals.
Capability inventory: The generation agents are explicitly instructed to use the Edit tool to apply changes to the codebase.
Sanitization: No sanitization or validation of the ingested code content is performed before it is processed by the LLM sub-agents.
[EXTERNAL_DOWNLOADS]: The skill utilizes npx eslint for verification in Phase 5. If the required ESLint packages or plugins (e.g., eslint-plugin-jsdoc) are not present in the local environment, npx may attempt to download them from the npm registry at runtime.
[COMMAND_EXECUTION]: The skill executes several shell commands for environment discovery and linting, including find, grep, ruff, and staticcheck. These commands are standard for the skill's intended purpose of analyzing and verifying a codebase.

ensure-docs