The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill documents methods for ingesting data from an external Redis database.
Ingestion points: Data is retrieved using r.get(), r.hgetall(), r.lrange(), and r.json().get() as shown in SKILL.md.
Boundary markers: No specific delimiters or warnings against embedded instructions are provided in the code examples.
Capability inventory: The documentation shows standard data operations but does not involve dynamic execution (e.g., eval or exec) of the retrieved data.
Sanitization: The provided snippets do not include data validation or sanitization routines.
[METADATA_POISONING]: The skill refers to 'Redis v8.0' and 'redis-py v8.0', which are future or non-existent stable versions compared to the current environment. This could cause confusion or lead to potential supply chain issues if users attempt to pin dependencies to these strings.
[COMMAND_EXECUTION]: Documentation includes examples of potentially destructive administrative commands.
Evidence: The skill lists FLUSHDB and a SCAN pattern for bulk deletion in SKILL.md, though these are paired with cautionary notes.
[EXTERNAL_DOWNLOADS]: The skill provides installation commands for legitimate and well-known Python packages.
Evidence: Mentions pip install redis and pip install django-redis in SKILL.md.

redis