The Agent Skills Directory

[SAFE]: The skill is purely instructional and facilitates the use of local documentation (glossaries) for naming conventions.
[INDIRECT_PROMPT_INJECTION]: The skill directs the agent to ingest content from docs/glossary.md and local glossary.md files (Ingestion points). No explicit boundary markers or delimiters are defined to separate glossary content from instructions (Boundary markers). The skill's capabilities are limited to influencing the agent's prose and identifier naming, with no access to high-risk operations such as subprocess execution, sensitive file writing, or network requests (Capability inventory). No sanitization or validation of the glossary content is specified (Sanitization). The finding is classified as safe because the capability inventory is minimal and aligned with the skill's primary purpose.

consult-glossary