The Agent Skills Directory

[COMMAND_EXECUTION]: The skill extensively uses the Bash tool to perform local and remote operations, including repository management, code modification, and CI log retrieval.
Specific tools executed include git, gh CLI, and a variety of language-specific test runners and build systems (e.g., cargo, npm, pytest, gradle, mix, dotnet).
Sub-agents are instructed to run local formatters and linters after installing dependencies.
[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from external sources and incorporates it into implementation instructions for parallel agents.
Ingestion points: Retrieves feature specifications from user-provided URLs via WebFetch, and fetches issue descriptions, project details, and comments from Linear and GitHub.
Boundary markers: Absent. The instructions do not define clear delimiters or provide instructions to ignore potentially malicious directions embedded in the external specs or issue comments.
Capability inventory: The skill possesses high-privilege capabilities including arbitrary command execution (Bash), repository manipulation (git, gh), and spawning additional agent tasks (Task).
Sanitization: None. Data from external specifications is summarized but not sanitized or escaped before being passed to sub-agents.
[EXTERNAL_DOWNLOADS]: The skill clones source code from various SDK repositories within the vendor's GitHub organization (getsentry).
It uses the gh repo clone command to download target repositories for modification and testing.

sdk-feature-implementation