The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses grep and awk to locate existing headers, titles, and descriptions within the repository (src/app, src/pages, content/). These tools are also used to analyze the generated copy for stock phrases and sentence length constraints. These are standard, low-risk operations for the stated purpose.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process potentially untrusted data from external sources to inform its copywriting decisions.
Ingestion points: Reads support tickets, sales-call notes, customer reviews, and forum threads from 'shared folders' and the repository.
Boundary markers: The instructions lack explicit boundary markers or directions to the agent to ignore instructions embedded within the ingested materials.
Capability inventory: The agent has the ability to read from and write to various files in the repository (markdown, translation files, components) and execute shell commands (grep, awk).
Sanitization: There is no mention of sanitizing or escaping the content pulled from reviews or tickets; in fact, the skill explicitly instructs the agent to 'quote them exactly'. This could lead to malicious content from an external source being inserted directly into the codebase or influencing the agent's behavior during the drafting process.

copywriting