The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to process and improve user-provided copywriting drafts, which are untrusted data ingestion points. Malicious instructions could potentially be embedded in the text to influence agent behavior.
Ingestion points: User-provided copy assets, landing page content, and positioning inputs (documented in SKILL.md).
Boundary markers: None explicitly defined in the workflow instructions to separate user data from agent instructions.
Capability inventory: Local script execution (copy-scorer.js) and generic file-read/write capabilities provided by the agent platform.
Sanitization: No sanitization or filtering of external content is described.
[COMMAND_EXECUTION]: The skill instructions involve running a local Node.js script (copy-scorer.js) to evaluate copy quality.
The script uses the fs module to read local files specified by the user via the --file flag.
The implementation is limited to static text analysis (readability indices and keyword matching) and does not perform network operations, spawn subprocesses, or use dynamic code evaluation like eval().

aaj-copywriting