The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to execute several local repository scripts using npm run, including add:bansos for file generation, validate:data, check, and lint for quality assurance, and build for site generation. These are standard operations for managing the project's data-driven architecture.
[EXTERNAL_DOWNLOADS]: The workflow requires the agent to access and verify external URLs, such as official program landing pages and claim links. This interaction is necessary for the skill's primary purpose of verifying the legitimacy of aid programs before they are added to the catalog.
[PROMPT_INJECTION]: The skill processes untrusted data from external websites to populate project files (index.json, README.md), creating a surface for indirect prompt injection.
Ingestion points: Official program sources and CTA links fetched from the web.
Boundary markers: Absent; the agent is not explicitly told to ignore instructions within the fetched data, though it is told to verify claims.
Capability inventory: Execution of shell commands (npm run), file system writes, and Git pull request operations.
Sanitization: The workflow relies on the agent's manual review of generated changes and the execution of validation scripts before finalizing submissions.

bansos-add-entry