The Agent Skills Directory

[COMMAND_EXECUTION]: The skill includes a shell script for performing daily version checks to ensure the suite is up to date.
Evidence: The script uses curl to fetch version information from the vendor domain foundry.thehorizonfoundry.com. The process is rate-limited using a local stamp file at ~/.claude/.foundry-version-checked and does not execute the server's response.
[INDIRECT_PROMPT_INJECTION]: The skill has a potential attack surface for indirect prompt injection because it reads and processes documentation files from the repository, which could contain untrusted input.
Ingestion points: The skill reads various documentation files such as README.md, PRODUCT.md, and CLAUDE.md to reconcile them with the source code.
Boundary markers: It employs a 'sanitization sweep' that greps for specific markers like (finding: and (assumed, to identify and handle internal content safely.
Capability inventory: The skill possesses capabilities for reading documentation files, writing to the filesystem (to update docs and presentation surfaces), and executing shell commands for versioning.
Sanitization: The skill uses a strict whitelist for files rendered on public hubs and applies a section-level projection model to curate and exclude private technical details from external surfaces.

document