The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to process untrusted data from existing codebases and external documentation, creating a surface for indirect prompt injection.
Ingestion points: The agent reads local repository files and fetches external documentation/evidence via sandbase_run (SKILL.md, references/sandbase-api-map.md).
Boundary markers: The instructions do not mandate the use of clear delimiters or instructions to ignore embedded commands within the data being processed.
Capability inventory: The skill has the capability to modify repository files ("implement... code documenter work") and execute external API capabilities via sandbase_run.
Sanitization: There is no mention of sanitizing, escaping, or filtering the content ingested from the codebase or external sources before it is used to generate documentation or code changes.
[EXTERNAL_DOWNLOADS]: The skill utilizes the sandbase_* toolset to interact with the vendor's external infrastructure for tasks such as discovering capabilities, inspecting schemas, and running model inference (references/sandbase-api-map.md).
Vendor integration: These tools are standard components of the author's platform and include mechanisms for cost estimation and user approval before execution.

code-documenter