The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from the repository (code changes, configuration, and API specifications) to infer behavior and generate documentation, which creates a surface for indirect prompt injection.
Ingestion points: As described in references/workflow.md, the agent reads git diff outputs, OpenAPI specifications (*.yml/*.yaml), application configuration files (application.yml, application.properties), and Java source code files.
Boundary markers: The instructions do not define clear delimiters or "ignore" markers to prevent the agent from being influenced by instructions embedded within the files it analyzes.
Capability inventory: The skill allows the agent to execute git diff commands and write new markdown files to the docs/features/ directory.
Sanitization: The workflow does not include steps for sanitizing or escaping content extracted from repository files before it is interpolated into the generated documentation or the agent's context.
[COMMAND_EXECUTION]: The skill requires the execution of local shell commands as part of its primary workflow.
Evidence: references/workflow.md explicitly instructs the agent to execute git diff master...HEAD --stat and full diff commands to analyze changes.

document-feature