The Agent Skills Directory

[PROMPT_INJECTION]: The skill is susceptible to Indirect Prompt Injection (Category 8) because it ingests untrusted data from web searches to determine configuration steps.
Ingestion points: External search results fetched via firecrawl_search in Phase 1 (SKILL.md).
Boundary markers: None identified. The skill does not use delimiters or instructions to ignore potential commands embedded in search results.
Capability inventory: The agent can execute package installations (npm install), write configuration files (Vite, Next.js configs), and create script files (Service Workers).
Sanitization: No sanitization or validation of the search-derived content is performed before it is used to generate code or commands.
[COMMAND_EXECUTION]: The skill executes shell commands for environment auditing and tool execution.
Evidence: Runs npx lighthouse in Phase 7 to perform PWA audits.
Evidence: Runs framework-specific build and installation commands (npm install).
[EXTERNAL_DOWNLOADS]: The skill downloads and installs external dependencies from public registries.
Evidence: Installs the @ducanh2912/next-pwa package from the npm registry in Phase 3.

enhance-pwa