The Agent Skills Directory

[DATA_EXPOSURE] (SAFE): The skill only accesses internal resource files (references/codes.md, references/formats.md) via a standard resource reading tool. No sensitive file paths, credentials, or environment variables are accessed.
[REMOTE_CODE_EXECUTION] (SAFE): No code execution patterns, subprocess calls, or dynamic evaluation techniques are present. The skill consists entirely of markdown instructions and data.
[EXTERNAL_DOWNLOADS] (SAFE): No external dependencies or remote script downloads are defined in the metadata or instructions.
[PROMPT_INJECTION] (SAFE): Instructions follow standard patterns for task definition. No attempts to override system prompts, bypass safety filters, or extract system instructions were detected.
[INDIRECT_PROMPT_INJECTION] (LOW):
Ingestion points: The skill accepts user-provided error codes as input for lookup.
Boundary markers: Not explicitly defined in the prompt interpolation logic.
Capability inventory: Uses read_skill_resource to read local files. No high-risk capabilities like network access or shell execution.
Sanitization: None observed, but the risk is mitigated as the lookup targets static, trusted local files rather than arbitrary paths.

reference-lookup